Catch up on stories from the past week (and beyond) at the Slashdot story archive


Forgot your password?
Linux Software

Dvorak On Linux And "The Big Time" 410

Cyberllama writes "John C. Dvorak's latest commentary is up at ZDnet. " I've been seeing this story came across quite a bit today. Dvorak offers an insightful commentary on Linux and "The Big Time" (He uses IRC servers as an example), although one that I don't necessarily agree with. In a switch from the normal take, he sees strong growth on the desktop, while predicting Linux won't grow in the workhorse server area much. Can anyone confirm or deny the allegation the IRC servers and Linux mentioned in the piece? Update: 09/21 06:29 by H :Click below - proof is below that Linux does at least some of the machines.

[root@brain:~]# queso -p 4400 * Linux 2.1.xx
[root@brain:~]# queso -p 6667 Haarlem.NL.EU.UnderNet.Org * Linux 2.1.xx

This discussion has been archived. No new comments can be posted.

Dvorak On Linux And "The Big Time"

Comments Filter:
  • ...according to Dvorak. Try telling that to HR people trying to find programmers to hire. It should be easy, according to Dvorak.

    The nature of the computer business is shitfting, however. Shifting away from shrink-wrapped software and into custom solutions, and open source.

    "Without the Net, the computer business would have been in the toilet years ago." What a stupid statement. That's like saying without railroads, the locomotive engine would have died out.

  • I used to hang out with the Sysadmins of the IRC servers on EFNet(Like 5 years ago), and I can tell you, they WERE BSD nuts. However they also wouldn't link Solaris server either(only SunOS) because Solaris had some sort of limit in their networking stack. I remember hearing over and over again about how bad Solaris was, just because of this one problem. If a person with a address came on-channel, they would spam and bombard them with comments like, "fix your sucky OS", etc.

    I agree that IRC servers take a pounding, but much of their pounding is due to poor initial design of the ServerToServer/Client protocols.(The coding is good, but you can only improve a protocol so much through code optimizations)

    IRC ops, especially EFNet and Undernet have very long memories and hold grudges like you wouldn't believe. :-) Petty-Power-Trippers comes to mind as a description. :-)
  • I always liked Dvorak and he makes a really good point here. This isn't your run-of-the-mill editorial here. I have Dvorak extraordinarily perceptive and in this case, he has done some research. And he made an important point here, even if it is one you didn't want to hear.

    Linux isn't great for all things. I fear that Windows 2000 will replace a lot of Linux servers in the future. So prepare to Wave 2 when Linux takes the desktop.

    Desktop? Yes. There are far more hackers with desktop machines than with server class machines to develop software with. So the desktop is natural. And I really doubt Microsoft will be able to make the convergance between NT and Windows/DOS systems smoothly enough for the regular user. And by then GNU systems will be able to do anything Windows can do now, and better!

    Notice I said GNU systems. In competition against propietary software, I really doubt Linux based GNU systems hold all the cards. Come on guys! we have 2 or 3 other very capable allies for the free kernal, why exclude them simply because they don't have the public eye? Software is much better at cooperating than competing, so why compete if you don't have to!


  • Blind advocacy such as this gets the Linux movement *nowhere*. A thread earlier tore into a FreeBSD user because he voiced his opinion that FreeBSD was a better server solution for him; rabid Linux advocates said they'd believe it if it were backed with facts. Well.. where are the Linux facts? Don't throw stones through your glass walls, ladies and gentlemen.
  • There was an alpha port of freebsd awhile ago.

    This is how the *BSDs get along.. they all develop in 3 directions then merge their progress.

    Anyhow, if you want portability run NetBSD.

  • by Pika ( 49094 ) on Tuesday September 21, 1999 @09:36AM (#1668901)
    Dvorak article is one sided, incomplete, unfounded, subjective, and seeks to destroy a 'technological-ideology' if you will.

    And you know what?? It sounds very, very familiar.... Maybe not his point, but definately the style.

    I know, I know!! It is the typical over-zealous linux supporter's statement, except reversed!! This time, the table has been reversed, and linux is getting unfair treatment!

    All the slashdotters read that article with their hearts nearly stopping, wondering how could anyone be soooo uncompassionate, and so blind at the same time.

    Really ... just think about it. 90% of linux related posts come from immature linux zealots who refuse to acknowledge any other OS for their strengths, blindly defending their choice of linux variant as the end-all of every computer related problem. Well, to those who this applies to, before you blast Dvorak for his stance, please re-evaluate the way you present your opinions. As it is now, you look just as stupid as he does.

    To the small minority who constructively defend Linux, while acknowledging its weaknesses, I commend you. If Dvorak was to have had a more open mind like yours, I know his article would have been much different, celebrating Linux for its true strengths.

    Those who can love linux and acknowledge its shortcomings are the only ones who will help us overcome them. Those who claim it's perfection and deny it's weaknesses are only hurting our cause, and not promoting the positive growth of Linux.

  • The FreeBSD developers wouldn't agree with your sentence. FreeBSD is not "necessarily a superset of linux"

    FreeBSD provides a syscall emulation layer which means most everyday Linux apps can run on FreeBSD.
    Since most everyday Linux apps are GPL or BSD, and therefore come with source, who even cares?

    IMHO The Linux development model is better, and the code for both is Free, so I support the one which matches my philosophy best (Linux)
    The only *BSD machines we have in this lab are testing IPv6, they seem fine but a bit old-school for my liking. Takes all sorts though.
  • eBay runs Solaris for web servers? I think not, my ignorant Sun advocate:

    sterwill@lister [~] telnet 80
    Connected to
    Escape character is '^]'.
    HEAD / HTTP/1.0

    HTTP/1.0 200 OK
    Server: Microsoft-IIS/3.0
    Date: Tue, 21 Sep 1999 21:41:02 GMT
    Content-Type: text/html
    Accept-Ranges: bytes
    Last-Modified: Tue, 21 Sep 1999 21:30:02 GMT
    Content-Length: 17437

    Connection closed by foreign host.

  • We assumed this might be so, but why so few IRC servers in the big nets?

    No, I can't spell!
    -"Run to that wall until I tell you to stop"
    (tagadum,tagadum,tagadum .... *CRUNCH*)
  • IBM's JVM is looking very promising. It's far better than Blackdown's, that much is certain. We're running the IBM JVM for our servlet-based apps with JServ 1.0/Apache 1.3.9
  • I assume by "effnet", you really mean EFnet (Eris Free Net). Usually the splits (not crashes) are attributed to network problems, rather than host problems. You can have the bestest(tm) box in the world, but it doesn't take that much effort for someone to fill your pipe.

    There are very few Linux IRC servers on EFnet. I know in the past, Linux has not performed well as a large scale IRC server. Many things have changed since then, and they probably have gotten better, but people still go with what they know. Seems to me FreeBSD is the flavor of choice for EFnet nowadays.

  • Looks liked John leaped before he looked.

    The Undernet's Central US routing server handles on average 1/3 of the Undernet population and is running Linux with kernel 2.2.11.

    And there are other server in the Undernet running on Linux, plus the preferred development platform now seems to be Linux.

    This is a long way from Never running Linux on the Undernet.
  • It seems that with the newer versions of Redhat, SuSE, Slackware, Debian, etc.. that they want to focus their marketing on personal desktops. Thus, that's where the code is being developed more. You don't see too many RedHat distro's default config'd for 4096 file descriptors. But I still think Linux has the potential, in most cases it needs some time to tweak and modify certain things. That's usually why BSD can handle it because you don't see too many people buying BSD distros at Best Buy like hotcakes.
  • Quick, rabid Linux advocate patrol to the rescue! Didn't you listen to Johnny? Say something bad about Linux, and somebody will pop up spouting fire and brimstone about how some other OS sucks worse. Grow up, pal. You're not doing your OS of choice any favors. Not to mention XWorld doesn't have the number of users compared to EFnet :)
  • I know for a fact that a good number of SlashNET IRC [] servers are running under a BSD unix. There are a smaller number that run Linux. The SlashNET server that's hosted here used to run Debian, but we switched it over to FreeBSD due to some security issues.

    Anyway, be sure to stop by SlashNET's #slashdot channel if you get a chance. It's fun, or something.

  • Windows:
    Download the EXE file, click on it
    Click Yes, Next, No, Proceed, Yes, Next, Next, OK
    Watch the entire machine reboot (huh?)

    Download the RPM file (not my fault if you choose to make your own life difficult and run Slackware)
    Double Click on the RPM file in your graphical file manager
    Ta da! No need to even logout (for simpler apps anyway)
  • Is a 386 with 16MB ram and 500MB hard drive (vast open expanses) low end enough? syslogd has problems keeping up with fetchmail/qmail, but otherwise it does it's job extremely well (firewall/ipmasq). I'm inclined to disagree that Linux has forgotten the low end, but it does work extremely well on modern hardware.

    Efficiency is Efficiency. If it works well on a pentium, it will work well on a 386, just a lot slower.

  • by Suydam ( 881 ) on Tuesday September 21, 1999 @06:46AM (#1668919) Homepage
    He's not entirely wrong. Linux isn't always the best task for the job.

    It's just that, his tone angers me. He takes the approach that Linux is A Bad Thing because it has weaknesses. That's absurd. Everything has weaknesses. Linux just overcomes its weaknesses quickly and efficiently.

    There's no need for taking an inflamatory tone when stating facts....he could have written the same article, state some of the same facts, and said merely "due to these shortcomings, there are times when Linux isn't the best solution." If he'd done that, I wouldn't be so angry.

    ....of cousre most of his articles have read like this one. Too bad.

  • Why should anyone start to believe him now?
  • ZDnet is wrong, people have known that for quite a while and thats why most people take it with a grain of salt.

    We run "truly intensive" applications here on Linux all the time. I think anybody would be hard pressed to find a more intensive application that a military combat simulation model with about 5000 entities. Linux chugs away and spews out the results we need when we are finished.

  • I'm in the process of nmap'ing a list of undernet servers, atleast 5 of them are running Linux.

    Thus far, we have:

    Those that aren't are, for the most part, either a BSD variant or Solaris 2.5+. There are many valid reasons why Linux doesn't have the same server share. I think Dvorak is intentionally inflamatory.

  • there should be three linux types: desktop, mid-range server, and highly scalable smp ... i think load-balancing is neat-o ; dvorak should check it out sometime...
  • He's speaking of Linux as if it's similar in scope, or attitude to Win32. As if Linux ships once every 3 years or something. If you obtained a brand new copy of Win95 in June 1995, and I obtained a brand new copy in August 1997, we most likely have two different products(Due to Rev. B and Rev. C). Linux changes at a much more rapid rate. When I went up from Kernel 2.0.36 to 2.2.9 there was a world of difference in speed and stability.

    While it's true that Linux can't be all things to all people, it can be more things to more people than any other OS out there. BTW, I'm nut a linux fanatic, I'm not really much of an advocate, I recommend the right tool for the right job.

    For some people NT is what they need. For others MacOS is what they need. For me it's a combination of 98/95/NT/MacOS/and linux.

    As far as Dvorak's fear of flaming, I'd say that it's unfounded. Those people who know the most about linux are the ones who are most likely to rip somebody a new one for spreading misinformation about linux.

    His comments are surprisingly similar to if someone mysogenist were making statements about women like "Women just can't hack it. Women are weak. Look at the evidence, women can't flap their arms and fly."

    Someone then responds "Well, men can't flap their arms and fly either. YOU can't flap your arms and fly."

    To which the mysogenist replies "That's not my point. Don't change the subject. I'm not talking about men, I'm not talking about me. You're personal attacks only prevent people from speaking the truth."

    Dvorak and people like him are talking out of their asses, and someone should let them know whan they're wrong.

  • by z4ce ( 67861 ) on Tuesday September 21, 1999 @06:51AM (#1668928)
    Well, I know _many_ small IRC servers do. Most IRC servers are running FreeBSD. But most developement for IRCd is done for FreeBSD. Posibly because awhile back in linux you had to apply patches to get many file descriptors. However, if you look through dalnet's old server descriptions you will find and ran linux. Both of those servers actually have quit dalnet since then. Albeit, I forget the name of the network they moved though. I'll see if I can find the URL for the server descriptions a little bit later.
  • by Anonymous Coward
    I have one word for John C. Dvorak: Deja []. This is of course the Site Formerly Known as DejaNews.

    Well, as it so happens, I work at Deja, and we use Linux for almost all of our servers, and believe me, Linux handles the (often quite heavy) load just fine.

    As it so happens, in our reception area, there is a 1997 award for "Technical Excellence" in the field of Web Design. Gee, can you guess who gave us that award? None other than our friend John C. Dvorak! It's signed by him and everything; it's a real hoot. I guess Linux was fine for "technical excellence" back then, but it has somehow become less so now.

    Someone else said it best; all he's doing is selling banner ads, not writing real or meaningful content. Not that there's anything wrong with that; you just have to go into it with your eyes open.

  • by diz ( 10034 ) on Tuesday September 21, 1999 @09:47AM (#1668930) Homepage Journal
    I was the one who originally ported ircII and ircd to Linux in 1992. Linux networking was just not stable enough at the time to maintain connections. I would lose my connections every few hours, so I ported everything to 386bsd 0.1 and everything was stable and rock solid. This is what originally gave Linux a bad rep in the IRC community. All the time I was running a 386bsd server, I was still developing for Linux though, because for all else, 386bsd was a dog.
  • I can't say that i've ever heard of any of the big three running linux, but i'm sure that they aren't running NT either.

    I don't think that NT could handle the load very well. And - he sites the fact that the IRC servers are constantly under attack, it *seems* that the NT TCP based buglists are a bit larger than that of most unixen. They are most likely running one of the major unix os's, HP/SUN/etc. In a few kernel versions, linux may be ready for that, but as far as samba/web/dekstop machines go, linux is OK for me.

  • Five years ago, Linux couldn't handle the load. Not even close. File-descriptors were only one problem.

    I would imagine that the "old school" effect also boils down to "Hey, there are a lot of shenanigans that go on in IRC land, and we've already proven X, Y, and Z work. The last time we tried Linux in any large way, it didn't work, so we're in no hurry to try it now."

    All I can say is that I'm happier in Dilbert Land as an engineer than I was dealing with the stress of working at an ISP. At least engineers tend to be a little more rational than alot of the bozos you meet online. :-)

  • by Shane ( 3950 ) on Tuesday September 21, 1999 @06:51AM (#1668933) Homepage
    Efnet and dalnet have few linux servers because the HUB servers refuse to link a Linux server. This has been the case for over 5 years. You have a couple choices. Use *bsd, use solaris or find a network that will link you.

    This has little to do with linux's current ability to perform as an IRC server and alot to do with the fact that HUB admins are "old school" and of the opinion that linux isn't UNIX and its a toy.

  • Now the best part is - the readers provide the content for nothing.

    umm, what content does /. have other than the comments? Not trashing the dot just saying...

    Anyway ZDNet's talkback is stupid, you can't even reply to and point out the boneheadedness of other's replies, /. definitely has the best talkback setup I've seen. (yes, I know it is modeled like Usenet, why change a good thing)
  • ok that's understandable. Tho, my K6-2 350 runs at 30-45C and is rated at 70C, though I haven't compared that with the average. (A friend of mine had an Intel Pentium MMX, _not_ overclocked, and the top of the heatsink was always too hot to touch...) Btw, it's worthwhile to note that since AMD chips are often used in cheaper systems, it's not uncommon to see substandard motherboards (probably not in your case) that often cause problems. Of course, I'm by no means saying that AMD chips are perfect, or even better than Intel's... I just really don't trust Intel, and I see too many people who do.
    Actually, what I really would like to see is average lifespan of Intel vs AMD vs Sparc vs Alpha vs etc chips when run 24/7 in server environments, but that may be too much to ask for. Seems the media is too concerned with benchmarking Linux vs NT to provide any really worthwhile information. Sigh.
  • I am not sure I agree, while it is pretty obvious that Linux will shine long and hard on our desktops, I see the biggest challenge and NT war on the small department layer.

    And there the answer is very simple, sure you have the databases vendor now taking first step but you don't really have the large choice of commercial software you can find on Solaris and NT. I am now contracting in a company where we are replacing NT by Linux with Jserv apache, in other words we developed under NT and now switch to Linux because it is in java.

    Servlets are pretty easy but large corporate app servers are missing a strong support of JVMs on Linux

    -- now pray and wait for IBM--
  • I'm sorry, but I've grown tired of Dvorak and his so called "knowledge" of the computer industry. Every time he attempts to discuss what "should be" or the way things "should work" or how something is going to "come out", I feel a sickening wave rising up from my stomach. And that ain't soup.

    If he ever really understood this business, it was a long time ago, most likely back before his scalp was sealed over with asphalt. Anyone reading Dvorak nowadays for anything more than insider tidbits deserves the searing pain in the temple such an activity elicits.
  • 2-5 years? Umm... The K6 hasn't been out for 5 years. It was originally released in April of 1997, not even 2.5 years ago. Also keep in mind that they went from .35 micron to .25 micron, which reduces temperature (and thus probably increases life expectancy) at similar clock rates, with the 233MHz+ K6, at the end of 1997. So to claim they last 2-5 years is kind of premature, especially when they changed the manufacturing process less than 2 years ago. You could say that they usually last 2 years, maybe, but not 2-5.
    Personally I'm happy with my K6-2/300, but it's only 8 months old, so I can't say much about the longevity of it.
  • He seems to toss aside the possibility of Linux taking on office desktops. Why? The graphical environments are much farther along than he gives credit, and products like Applixware, Star Office, and soon Corel Office Suite will give users real choice for office software.

    Yes, linux will be present in embedded or low cost machines (empeg car) and your local mom 'n' pop ISP, but there's no good reason it can't take over at the office.
  • If we wanted another Windows then you are correct. I personally want a home OS that does not crash.
  • If you know how to properly implement server clustering, you can run quite large web sites with Linux that uses the 2.2.x kernel.

    Fortunately, high-end server hardware are coming down in price, and a small "wall" of Dell PowerEdge servers running S.u.S.E. or Red Hat Linux with server clustering can even handle the front-end access to eBay. The nice thing is that by using a whole roomful of smaller servers, you can have server mirroring and other advantages of running many servers in parallel, so if one server goes down, shut down offending unit, replace with identical unit, and have the clustering software run an automatic rebuild process so we don't skip a beat! (^_^)
  • by Anonymous Coward on Tuesday September 21, 1999 @01:36PM (#1668944)

    I code for DALnet, and yes - we dont like Linux on our servers, for a number of reasons I'll touch on in a sec..

    First of all, I've seen a few posts to the effect that IRC doesnt eat much system resources. Allow me to be blunt - thats a load of crap. Our present version of IRCd (Bahamut) eats resources like a bitch, and it outperforms our old IRCd (Dreamforge) by about three times. Allow me to show you some stats:

    Random DALnet server:
    Current local users: 2081
    35795 ircd 2 0 51768K 48976K select 161:22 47.95% 47.95% ircd

    This is on a PII 400. The idea that Client servers dont require as many system resources as hubs is an uneducated assumption. Hub servers use substantially less CPU (and FD's for that matter) than client servers. They do, however, use slightly more RAM for buffering data flow.

    One quick thing on the IRC protocol - most large networks are using TS3 at this point, which includes a few bandwidth-saving perks (such as SJOIN).. the protocol handles the ~40K clients on DALnet in about an 8MB resync.

    Anyway, back to the topic at hand. As far as i know, there are no DALnet servers running Linux. Most are running FreeBSD, and we have a few BSDi and Sun/Solaris boxes. Linux doesnt work for our client load for a few reasons:
    - We need FD's. 256 (or 1024) just arent enough. And patching is a real pain.
    - Security issues. Someone mentioned the fact that pre-2.0.31 kernels wont be linked. Remember teardrop?
    - I've personally enjoyed the fun of occationally needing to hack sections of the IRCd to get it to compile - although this is rare, and no exception for other OS's.

    I develop under FreeBSD. We work mostly under FreeBSD. Linux doesnt do exactly what we want. The same is true to a company that wants software X developed on and for SCO or something equivilent. Personally I think that Linux is an underpowered OS, but I havent played with it in ages.

    One little sidenote - if there are any linux developers reading this - PLEASE make FD's a configurable option in kernel compile. Then maybe someone could easily stick up an IRCd and run up the clientload a few thousand and see how it performs. Until then, I refuse to expend extra energy looking into a different OS.

    Note: My opinions. Not DALnets.

  • by Beethoven ( 9076 ) on Tuesday September 21, 1999 @09:54AM (#1668955) Homepage
    No matter how trite or irrelevant Mr. Perens' comments may be, they will always get moderated up as insightful or interesting.

    Perhaps moderator mode should hide posters' identity.

  • He could have done what you suggest and stated the facts clearly and succinctly without spewing vitriol and digital vomit everywhere in that article.

    But then would it have gotten linked here on /.? Imagine how many hits ZDnet's getting right about now. The money from those ad banners is *pouring* in.

    - A.P.

    "One World, one Web, one Program" - Microsoft promotional ad

  • Well, it's not just ads from Microsoft, but every other Windows software vendor that wants to take advantage of the Win2000 launch and the opportunity that that creates to get your product on Win2000 desktops. It's in ZD's best interest to hype up Win2000 as much as possible because that's what puts bread on their table (at least in the short term). At least much more bread than Linux ads would.

    They only way for Linux to combat this is to have a thriving commercial software market that buys lots of ads,

    Six months ago, this wasn't true. They wrote alot about Linux simply because writing about some pre-beta of Win2000 doesn't sell too many magazines. (As a reference, go back and look at the OS/2 coverage during the Win95 waiting period.)

    Ironically, as Microsoft monopolizes more and more of the client software market, the number of advertisers and advertisements starts to fall. For example, there probably were many more office suite ads back in the day when WordPerfect and Lotus had a chance. Of course, as competitors fall, Microsoft ends up buying an increasing percentage of ad pages, giving them more leverage.

  • by Anonymous Coward on Tuesday September 21, 1999 @02:08PM (#1668973)
    In general, FreeBSD is better for me than the Linux distributions I've used because it has an integrated source tree for the entire BSD environment (kernel and user). That means there's a single source tree for the bulk of what I use, and that the bulk of what I use is developed (or at least ported), maintained and distributed by the same people. Even where I have to use external code, the `ports' tree provides a mechanism by which it can be maintained, scrutinised and approved by FreeBSD contributors.

    The bottom line for me is that the standard FreeBSD distribution provides a full, traditional UNIX environment, without all the unwelcome additions Linux distributions seem to add. My experience with Linux distributions was that they took up much more disc space, but provided much less that was actually useful to me.

    From a technical perspective, FreeBSD is better for me in the following respects:

    VM: On my systems, the performance of the FreeBSD virtual memory system under load has been much better (more responsive, better throughput).

    Filesystem: I've run into occasional filesystem corruption on Linux, owing to its default use of asynchronous writes for metadata. FreeBSD doesn't use asynchronous writes for metadata by default, and its `soft updates' scheme provides a completely safe mechanism for writing metadata asyncronously (as opposed to the blind approach Linux takes, which can leave the filesystem in an unusable state).

    Filesystem Layout: The Linux distributions I used had quite different filesystem layouts, all of which differed from traditional BSD and System V layouts. FreeBSD (and NetBSD/OpenBSD) uses the BSD filesystem layout, which has been around for a long time, and is fairly good (logical, consistent). FreeBSD also uses /usr/local for external files installed via the `ports' tree, rather than dumping them into the standard tree (shudder).

    TCP/IP: The TCP/IP implementation in BSD is the reference implementation, so its behaviour is better known and documented. It's also supposedly faster, although TCP/IP bottlenecks are not something I run into (I tend to run into memory and network-bandwidth bottlenecks), so I've not noticed a difference.

    From a technical perspective, Linux was better for me in the following respects:

    Hardware support: I've got one perhipheral (by a defunct manufacturer) that won't currently run under FreeBSD. I can get by without it, but I miss it.

    On balance, FreeBSD's advantages make it much more useful for me. For you, it sounds like the reverse is the case. If, however, support for your hardware is added, you may want to give FreeBSD another chance.
  • Check out html [].

    The Dvorak keyboard is an alternative layout, not a superior one.

  • It appears that both of your Anonymous Coward and named versions of your comments got marked as flamebait. Before you cry "Foul! Unfair Moderation!", why don't you think about what you are posting:

    How long until linux users just bite the bullet and admit its better?
    That is a groundless insult to the Linux users. If you are going to say X is better than Y, give a reason. An old DALNet post referring to an ancient version of Linux is not a reason it's bad, or should I disrespect all BSD users because of licensing problems with NET-2. It's just as ludicrous.

    FreeBSD software is necessarily a superset of linux.
    I don't think anyone would agree with that one.

    So it's either flamebait or a troll, live with it.

  • Try scanning the irc ports like so:

    dnsquery | awk '/IN A/{print $5}' | xargs queso -p 6667 * Standard: Solaris 2.x, Linux 2.1.???, MacOS * Dead Host, Firewalled Port or Unassigned IP * Standard: Solaris 2.x, Linux 2.1.???, MacOS * BSDi 3.0, IBM S/390 * Standard: Solaris 2.x, Linux 2.1.???, MacOS * Linux 1.3.xx, 2.0.0 to 2.0.34 *- Unknown OS, pleez update /usr/local/etc/queso.conf * Linux 1.2.xx * BSDi 3.0, IBM S/390 * Linux 1.2.xx * Standard: Solaris 2.x, Linux 2.1.???, MacOS * Linux 1.2.xx *- Unknown OS, pleez update /usr/local/etc/queso.conf * Cisco 11.2(10a), HP/3000 DTC, BayStack Switch * Linux 1.3.xx, 2.0.0 to 2.0.34 * Standard: Solaris 2.x, Linux 2.1.???, MacOS * Linux 1.2.xx * Linux 1.2.xx *- Not Listen, try another port *- Firewalled host/port or network congestion *- Not Listen, try another port
  • Not only wasn't 1.2.x good enough, 2.0.x wasn't either. It required patches just to get a decent number of file descriptors available ( was having some problems with this running the 2.0.36 kernel). Hence the various comments that Linux is not good enough for an IRC server. I'm not sure if 2.2.x improves things, but I hope so.
  • No serious site is going to want someone posting uninformed flamebait all the time (except maybe /., I'm still here after all) because every time Dvorak pisses us off he turns us off from PCMag. Everytime he prints something so blatantly phony he reminds us about the shallow level of analysis on PCMags site.

    He is not getting it any closer to being a site I regularly visit. Which is what he actually should be doing to keep his editors happy...


    They do not care about YOU! YOU are far too technically astute. YOU are not ZD's market! Nor are you M$ market! I posted something to this effect quite some time ago on ZDNN. Althought I was writing about M$, it is equally applicable to ZDNN. Rather than post a link to it on ZDNN and thus provide more grist for their banner serving mill, I'll just quote it verbatim:

    I have come to the conclusion that MS isn't stupid, its just that MS has realized that people like you, myself and many of the other Talkback posters here don't matter. We are laboring under the false impression that it's people like us they want to sell software to. It's not! All the money of all the halfway knowlegeable/informed/technically astute folks out there is a drop in the bucket compared to the trillions of readily available dollars in the hands of those who know no better. The 11 million people who think AOL is the internet. The ones who still use Netscape 1.1, with set as the home page, because thats what they got when they signed up with their ISP. People who dont know what drag and drop is. People running their $600 17' monitors in 640 x 480 mode, with Start buttons the size of my shoe. We simply don't matter. THOSE people are the great untapped market of future software sales - they will buy anything you tell them they need, and when it makes a mess of their system, they'll beleive it's their fault, because they readily beleive Bill Gates invented the internet, and TCI invented cable television, and that both televisions and computers are smarter than they are...

    Bite My Ziff, Davis!

    "Cyberspace scared me so bad I downloaded in my pants." --- Buddy Jellison

  • yep, the biggest OS on EFnet used to be SunOS, then Solaris, then FreeBSD. Linux has in the past had trouble handling large numbers of open fd's, which is the one thing that ircd stresses in an OS. up to some very recent versions, the stock kernel just didnt let any process have thousands of fd's; vendor kernels like redhat have shipped with large-fd patches for a while (rh6 has the hard limits for fd's set to 1024, and the kernel should be able to handle even more). I *think* that the large-fd patches have finally made it into the mainstream kernel, so the latest 2.2 should work noticeably better.

    other than that, well, nothing wrong with using the best tool for the job, and FreeBSD has always been good at this, so there's nothing wrong with using FreeBSD for irc servers. but Linux is not "crippled" in that respect, and it makes even less sense to say that Linux "will not make it in the server area" because of that. How many "enterprise" servers out there run large ircd's, or software that pushes the OS in the same ways ircd does?

    btw, there *have* been Linux servers on EFnet. i'm not up to date with what's going on there now, but I know used to run Linux.

  • I sort of disagree. I don't think the order that articles appear should be based on moderation. Then too many people would be complaning that articles they wanted to read where dissapearing (even if you could turn it off in settings) though having a little score box next to it would be cool. Though on the thought of the order of articles I have always thought it would be cool to give articles a boyancy rating. Of course their initial boyancy would be simply based upon age. The newest articles would appear on top. But articles that recieved alot of comments, would start to float up to the top, thus the people would continue to discuss this article till it became boring and eventually sunk into the depths of article decay. (maby have something like a article starts out with a boyancy of 0 (zero meaning it floats to top) and as it becomes old it gains say a point an hour (or whatever works best) but for every 50 or 100 post (this would be a variable that would need tweaking) the comment losses 1 boyancy point. So as long as a comment is receiving talk it stays ontop (though you might want to make sure some how that brand new articles stay ontop for atleast a half hour or so to give them time to catch on and have people start commenting)
  • I don't know much about IRC and big-time servers, so I won't comment on that. I do think that his conclusion that Linux will shine in the low-end desktop market is a bit far-fetched. Here's my thinking.

    In order to have a good desktop system that my mother and grandmother can use, you're gonna need a GUI that allows drag-and-drop, cut-and-paste, standardized widgets and stuff ABSOLUTELY EVERYWHERE. I think the only way to do this is to have these features built into the X Windows. GNOME and KDE just won't cut it. Furthermore, there should be some way to force software makers into supporting this native desktop environment.

    You're also going to have to do something with fonts and printing. There should be one centralized place to dump your fonts, and any font that X can display should work with your printer too. Of course, TrueType fonts should be natively supported.

    Device support is still a major concern. Something should be done to provide a standard software interface for printers, modems, scanners, digital cameras, and the like. Scanners are pretty cheap these days, and digital cameras are coming down in price too, so it's possible that a low-end computer would need to support them in the near future. Even cheap-o devices (in fact, especially the cheap-o devices) should work reliably with Linux.

    These are just some of the problems that need to be worked out before Dvorak's predicition of Linux on the desktop can be taken seriously.

    Gotta go eat dinner now. Take care,


  • Cripes, Solaris has been in development for how many years now? With how many developers?

    Linux is still basically a PC program that has been ported. PC hardware isn't designed to take this kind of load. The hardware itself can't keep up with it. Linux is not *designed* to be a massive server. It's meant to be a desktop for serious technoid users who like technical excellence in their home machines. It happens to make a good server because it's good software, not because it's really designed to be one.

    The fact that it trounces NT so thoroughly isn't really an endorsement of Linux, it's simply stinging criticism of NT. Linux wasn't written to be a server: the fact that it is a better 'server' than NT (which IS meant to be a server) simply shows how low the MS standards have gotten.

    The underlying technology is being worked on. Each year that goes by will see it get stronger and stronger. Someday it may run Ebay and IRC servers... but then again, it might not. And that is okay. Ebay is gravy.

    There is a *reason* Solaris boxes cost $50K plus. And no, the reason isn't (just) that they are greedy. The hardware is designed for an entirely different scale of problem. Sun knows how to make servers, and how to approach the problem from the bottom up. Linux is an accidental server, pressed into use because it happens to be better than the only commercial alternative that most people can afford.

    You might as well bash the VW Bug because it isn't as fast as a Ferrari and doesn't pull the load a Mack truck would.

  • The previous poster said that FreeBSD would not run on his or her Alpha.

    That might be correct, depending on which Alpha he or she referred to.

    Some systems (often marketed as NT/Alpha) use a boot system called ARC which isn't currently supported in FreeBSD. Thus, you cannot boot FreeBSD on these machines which renders the port virtually unuseable.

    Btw, NetBSD won't run on these systems as well. From irmware

    Can NetBSD/alpha be run on systems with only NT (ARC) firmware? (top)

    Not currently. NetBSD/alpha requires the SRM Console firmware (used by OSF/1 and OpenVMS) to function. There are two main reasons for this: the console
    software is what's responsible for loading the operating system's primary bootstrap program, but the NT firmware's method of doing this is undocumented; and
    the firmware provides the PALcode for the system, which handles low-level memory management and interrupt handling details on a system-specific basis.
    (NetBSD uses the Digital Unix PALcode.)

    The Alpha port of Linux, can boot on systems with NT firmware not because they can use the NT PALcode, but because MILO, the Alpha Linux loader, includes
    its own PALcode. (This of course means you need a different loader for each system type.) The PALcode they use is based on older Digital Unix PALcode and is
    missing a few functions. There is a possibility that this could be used for NetBSD, assuming someone wizardly enough wants to do the work.
  • The difficulty stems from two points.

    1. Linux was insecure and unmanaged.

    2. Linux only supported 256 FD's, and later 1024 FD's.

    The insecurity of linux in the begining was well known, it was a hackers OS. Plus, you'd have a lot of unknowns. No REAL packaging system besides tar files, it was a hit or miss proposition based on your admin.

    However, File Descriptors (FD) is one of the major reasons. For all of the 1.x kernels, you only had 256 file descriptors. and it was a pain to hack in more. when 2.0 came out, I believe you were still limited to 256, but it was a bit easier to put more in. Some of the later 2.1.x kernels allowed 1024 by default, which 2.2.x does as well. Anyway, other OS's, like Solaris, default to unlimited soft FD's. (hard FD's are still at 256 though). For every IRC connection, you need a FD. So... by using linux, you're automatically limited to 256 or 1024 people. Even if you hack in more, you still have a hardcoded limit. Once again, it's the admin that makes the difference. and as the major irc nets take off, they want something more substantial than 'I heard this admin is good.'

    Anyway, I base this on the fact that I used to run (EFnet) on an AIX machine. We could have ran it on a linux box, but it was just easier with the constant kernel thrash happening on linux to keep it on the AIX box. I now run [] (and used to run [] before it died) and that is on a linux box. and it was a pain to hack in the 4096 FD's we currently have. hence, I would have prefered Solaris at some points. However, Apache was much easier to setup on there, even though that required a lot more FD's as well. You have IRC nets all over the place that use Linux, it's just the history that stop it on the huge networks.

    As far as the total article, I see things opposite, of less linux desktops and more linux servers. If you've seen the things that Cobalt [] is doing, you'd see what a server can do with a microkernel. But the business side of linux just isn't up to speed yet, so it will stay in the hands of geeks and out of the hands of biz guys. which is fine by me.

    Gonzo "GoNINzo" Granzeau
    Gonzo Granzeau

  • That depends very much on your definition of ``runs on Alpha.''

    FreeBSD and NetBSD run on almost all Alphas, but not on those which use ARC to boot. See my previous comment for more information.
  • it runs all linux binaries (including civ and the like) after you add COMPAT_LINUX to your kernel and install the linux_base package.

    Funny, I did that and it still didn't run everything reliably. What it did run was pretty amazing, but I'm disputing the word "everything" here.

    it's not harder to configure than linux, it's _different_. it's not harder to customize than linux, it's _different_.

    For me, maybe I'm weird, it was harder. Period. I'm not lying, I'm an honest guy. It really, truly was harder. Especially kernel configuration, even with a FreeBSD guru hanging over my shoulder.

    i've never found it to be less reliable than linux, i'm not sure where you got that from.

    As I said in my post... *I got that from USING it.* For weeks. When I said I tried FreeBSD, I didn't mean I installed it, booted it, and said "I'm a mindless Linux freak. I'm going to reboot now." I mean I tried it.

    X works better for me on FreeBSD than it did on linux. KDE works better for me on FreeBSD than it did on linux.
    Emphasis added by me

    That's great it works better for you, but that wasn't my experience. It was slower, and actually locked up and crashed several times. Granted, the whole system didn't crash. I could telnet in, kill X, and continue as if nothing had happened. But it shouldn't have crashed to begin with.

    kde 1.1.2 and x 3.3.5 are in ports, and i can compile and install them with my eyes closed.

    This is a recent addition. They were not there when last I tried FreeBSD (when 3.2 had been out for a short while).

    you did not compile and/or install your new kernel correctly.

    To be blunt, you are wrong. I did. I had the help of someone who only uses FreeBSD, who has since version 1.something. He didn't understand the problem. This was with version 3.1. Following the same procedure with 3.2 resulted in perfect success. So, musta been a bug they fixed. But speaking about the kernel, it's really tough to configure. I can do it, no problem, but it takes too long. Yes, I can tell you are the kind of person who believes vi is the only tool you ever need, and I can see that, I can do that, but it's a waste of my time.

    FreeBSD 4.0 is in development with egcs 2.95.1, the developers #1 priority is to stability, not bleeding-edge software.

    First, it isn't egcs anymore. As of version 2.95, it's the official gcc. And OK, fine, stability not bleeding-edge software. What about gcc 2.8? Come on! There are important compiler features not supported at all by the default FreeBSD!

    You can install egcs 2.95.1 from ports and then edit /etc/make.conf and change CC = /path/to/egcs and (in my experience) it works fine.

    Except, as I said, it isn't egcs. It's gcc. And what about pgcc? Maybe that's why FreeBSD was so much slower on my Celeron than Linux, because I've recompiled everything on my Linux box with pgcc? I have a suscpicion it was pgcc that broke the FreeBSD things, but it doesn't break anything under my install of Linux (SuSE 6.1). I put emphasis there, because other people may tell other tales.

    so for all your real, actually experienced reasons, you're wrong.

    Except... I'm not wrong. I experienced this stuff, you know. Are you saying I was delusioned? Perhaps I was hallucinating? Come on, how can I be wrong, when all I'm doing is recounting my experience? It happened to me, not to you, and I don't appreciate you passing judgement on it. What you're saying is that my opinion isn't valid. That's no way to communicate with someone. I was simply saying what my experience was, and why I made the choices I did. You have provided no evidence to the contrary... for me. You say it's not harder, it's different. Well, gee, I found it's differentness to be more difficult. And now I'll let you in on a little secret: I started out on FreeBSD. I used FreeBSD exclusively (I don't play games) for a year. Then I tried Linux. I liked it better. And if you tell me I don't have the right to choose what I like, then... well, it's not appropriate for a public forum.

    Robert Thompson
  • by MuyJuan ( 9379 ) on Tuesday September 21, 1999 @06:55AM (#1669045)
    Every few months, you see an article discussing how Linux is not good enough for ____, which purpose is rapidly increasing. A year ago, Linux wasn't good enough for the desktop. Now, it's not good enough to power EBay (is that really true? He implies that it is). What won't it be good enough for tomorrow?

    As an aside, I notice that Dvorak likes to accuse Linux advocates of being groudlessly optimistic. I in turn find him to be groundlessly pessimistic. Linux might not be the answer to every possible need, but it's more likely to achieve that end than any piece of bloatware that microsoft is likely to put out.

  • by Amphigory ( 2375 ) on Tuesday September 21, 1999 @06:56AM (#1669049) Homepage
    Something I've observed is that some of the "big time networking guru's" that Dvorak alludes to have been around too long! That is, they've been around so long that the first time they tried Linux is WAS a buggy system that couldn't stand up under high loads.

    Many of them promptly went to *BSD for the market segment that Linux is targeted at. Also, I've got to say that BSD does seem to be preferred for MUD's, which are from a programming perspective very similar to IRC.

    However, this doesn't really validate Dvorak's conclusion that Linux will not be able to compete in the server market. (I've been reading his stuff for ten years and still haven't seen him be right). IRC is a very specialized application that you don't see much of in the "real world".

    I do think that Dvorak is right about Linux's big area of growth being the client -- there are just more of them out there. However, I think he's missing the boat calling for the low end client: until something is done about netscape and staroffice being pigs Linux doesn't run as well on low power machines as does windows '95.

    The real need I see for Linux is more tightly focused distributions. I don't think every distro should try to be both a client and a server. This would result in much cleaner installs for both servers and much faster, more robust clients.

  • by paul.dunne ( 5922 ) on Tuesday September 21, 1999 @06:58AM (#1669054)
    I saw this "they aren't running NT either" comment a few times on ZDNet, too. Why do people insist on doing this? You're only proving one of Dvorak's points for him -- that is, when you criticise Linux, don't expect a reasoned response. Too many people do this. "Linux can't do X". "Ah, neither can NT, yah, booh, sucks to you!". It would be both more interesting and better for Linux to hear some explanation of Dvorak's assertion about Linux not being able to run an IRC server: whether it's true, are there any examples of heavily-loaded Linux IRC servers, etc. We do ourselves a disfavour by assuming that anyone criticising Linux has ulterior motives -- promoting NT, for instance, as many people seem to think.
  • by richnut ( 15117 ) on Tuesday September 21, 1999 @06:59AM (#1669060)
    IIRC, some of the old DALNet servers were Linux when it was just a feldgling offshoot that was considered a joke by the Undernet/EFNet crowd. It's been a real long time since I gave a care about IRC archetecture though, so I could be wrong.

    The fact that IRC is such a ridiculous hog is becasue the IRC protocol was really never intended for tens of thousands of users and thousands of channels. AFAIK they still require every server to know about what every user is doing. That in itself is sort of ridiculous for a system that is supposed to provide global chat, but you'll find no crowd, ANYWHERE, more unruly than the IRC folks when it comes to change.

  • by CaptSwifty ( 61835 ) on Tuesday September 21, 1999 @06:59AM (#1669065)
    "The fact is, Linux has yet to prove itself at the top of the food chain, and until it can run eBay, for example, it will remain the "in-between" OS."

    Umm, the last time I checked, eBay didn't stay up all that much. Maybe they should try linux.
  • by Booker ( 6173 ) on Tuesday September 21, 1999 @09:19AM (#1669066) Homepage
    I found this page [] detailing some votes on new servers for Each proposed new server has information on the machine (including the OS), whether it was voted up or down, and why. Most Linux servers were rejected, with this sort of rationale:
    I think should not be linked to DALnet because: 1. Dual admin.. especially the bs co-admin who just owns the machine 2. Single T1 3. Machine runs linux
    but this one [] was accepted. However, the voters [] had several concerns:

    "To summarize, my only real problem with this machine is that it is running Linux and our own experience has proved to us that Linux is not stable enough to handle a large client load, and we are not the only ones who have experienced this same problem."
    "Although I would like to add that the fact they have and use Linux for the server worries me. ;)"

    This was in 1998, I believe.
  • by Anonymous Coward
    I think that Mr. Dvorak has an excellent point. In many cases, Linux cannot handle the huge workloads, but does that mean that it won't ever be able to? No, absolutely not. Linux development will continue to grow and emerge, so as far as the non-server argument, I think he may be wrong. Where he is right is in the idea that companies think they can't directly attack Linux because of all of the nuts out there that would freak out if someone said something bad about the holy OS. Hopefully, Linux users and Windows users, along with the more powerful UNIX variants will form a nice cohabitation that they all benefit from. But then again, that's just a dream, right? Where windows children and linux children play together....
  • by Anonymous Coward on Tuesday September 21, 1999 @06:59AM (#1669071)

    The DalNet Server application has this to say about Linux:
    Linux has been found not to be workable with DALnet's traffic load. All DALnet servers running Linux have been switched to FreeBSD. Therefore, if you are currently running Linux on the server for which you are applying, we suggest you switch to FreeBSD as soon as possible, preferably before sending in your application. Servers running pre-2.0.31 Linux kernels cannot be linked to DALnet at this time.
    The question is... how long has this statement been in there... and is it still true?
  • Is there a place to purtchess dvorak keyboards? Seems the only other alternative is to rip the keys off one and rearrange them and put super glue on them to get them to stick.

    Here [] is a perfect place to start shopping for dvorak keyboards, from the cheap to the heavy duty spring click keyboards. Get one with a built in stick pointer too!
  • by Anonymous Coward
    the inadequate-support explanation creeps in for no other reason than to keep the lunatic fringe of the Linux movement from clogging the e-mail system with complaints.

    Very very true. I've been flamed countless times for saying I like Windows more than Linux. You say anything bad about Linux, no matter the basis of it (be it flame or actually a good point), people will flame you. Maybe not every Linux user, but the fringe radicals spoil it for everyone. As Dvorak says, "Just hating Microsoft is not a good enough reason to promote Linux above everything else."
  • by CyberSnyder ( 8122 ) on Tuesday September 21, 1999 @06:59AM (#1669095)
    Truthfully, I think he was taunting the slashdot crowd to come on over and post your flames. The more feedbacks they get the higher their ratings. Just like the Neilsen ratings. Show alot of swimsuits, flame Linux -- same thing.

    Just an opinion.
  • I wish just one claim in that entire article had been backed up with more than an assertion.

    "Linux can't cut it--period. Why is this never mentioned by the Linux proponents who flood
    the critics with e-mail anytime anyone says
    anything to disparage the OS?" - Maybe because they'd feel compelled to back up such a claim?

    "The modem-equipped machine was good for
    remote access; then the Web was invented, and
    now there is nothing left to conquer. For at least
    a decade, the only thing that's been going on is
    the debugging of old code. Without the Net, the
    computer business would have been in the toilet
    years ago." - Mmm. Nothing like the taste of speculative fiction. I mean, this might be a good point - but why should I even listen to it?

    "On a $199 machine, the OS can't cost more
    than a few bucks." - I found something resembling a fact! Do I win a prize?

    He claims that Linux can't handle the strain of running a truly intensive application, and offers not an example of Linux having _failed_, but of two applications, each with a handful of options, and the fact that none of those options is Unix. Obviously, PSOS is the only viable embedded OS because _every_ intelligent battery plant manager in the Telecommunications industry relies on PSOS! (Not that there's more than three intelligent battery plant managers available...)
  • by stevew ( 4845 ) on Tuesday September 21, 1999 @07:01AM (#1669103) Journal
    Dvorak is in the business of generating hits for their website. He's infamous for creating flame bait. See the stuff from a few weeks ago relating to his comments about the Ibook (calling it girlie...)

    I could choose to dispute his claims concerning
    IRC, others have done that already. Just look at the style of the article -it's completely incoherent. First - Linux can't be good because it doesn't run IRC, then he babbles about no standard choice for X environments (how that relates to IRC I haven't a clue), etc. It's just a strung-together bunch of inuendo and half-truths with no significant research behind it.

    Just ignore him - we've all got more productive things to be doing - like writting useful free code!
  • Wow, the Sun apologist crowds here are thick (in more ways than one). You even found a way to work Linux into the architecture of eBay's commerce business, where, to my knowledge, it doesn't even exist! Please, next time you fire up the small grey lump of flesh in-between your ears, check the fluids and let it warm up.

    eBay's problems are many: management who isn't putting money in the right places, a lack of emphasis on redundancy in their architecture, and a poor choice of platforms for parts of their business.

    Their web servers are slow; they often refuse connections when they're too busy. In fact, they schedule daily down-time for maintenance, something unheard of in the Real World of e-commerce. These are computers, they don't need breaks.

    Sun makes some excellent hardware, but you can make it inefficient if you try hard enough (you probably have experience in this endeavor).

    That eBay is perceived as a company seriously in need of help implies that people think it has problems. Since Sun solutions never fail (as I'm sure you know), the blame must lie somewhere else, hence my citation of the _actual_ _problem_. The original poster in this thread took his opportunity to slam Linux's scalability when it wasn't relevant anyway. He isn't ignorant because he likes Sun solutions, he's simply ignoring the real problem for a chance to plug his favorite purple logo.

  • Dvorak hit the mark pretty well with his article. If it seems like he's baiting linux users, you're right. If you act suprised ytou need to kick yourself. Dvorak isn't a Microsoft agent as many people claim, and it's not FUD about linux, it's real world observations. You can't go around touting your opinions like they're fact. This reply is opinion (hence I think...) and I don't expect it to be taken factually. Linux is a very stable and well designed system and works great in the medium server class but because it was designed for PCs (thats Personal Computers) from the beginning trying to patch it onto a huge server probably is going to be alot harder than using something like a BSD. On the otherhand using FreeBSD as your desktop is a bit of a waste of computing power since Unix was originally designed for very large computers, BSD being able to work on such large computers. FreeBSD had the advantage of high-end knowlege but it also had the knowlege to be able to get it down to a desktop. Linux has had to start from the small PC and work it's way up to the huge high end servers. I highly doubt the Capt. Kernel Torvalds will ever try to get it up to the level of Solaris or AIX, at least not with Linux itself. Don't badmouth Dvorak because he hit a nerve, thats what he's paid to do.
  • If linux is no good on high-end servers as Dvorak states, then why has SGI elected to use a modified version of Redhat 6.0 as the new OS for it's servers? (
    While I am somewhat regretful that they are not carrying IRIX on to their new Intel based systems, I am glad that they have chosen to embrace linux over a Microsoft product.
  • There is actually a very good reason that Linux has not traditionally been used on IRC servers. However, it is possible the reason is outdated.

    The reason has to do with IRC servers needing one file descriptor for every user connecting to the server (along with a few others for miscellaneous purposes). Up until the 2.2.x kernel, Linux required extensive patching to handle more than 256 fd's per process. Not only that, but these patches were not well tested, nor were they great patches. While they allowed for additional file descriptors, they opened up numerous problems in that fact that the rest of the kernel wasn't ready to handle that many, especially not efficently.

    I've followed and been involved with DALnet since late '95, and that's the reason there have been few (Note, 'few'. There have been some.) Linux IRC servers. Under low load (Read: When using a number of file descriptors that doesn't require the kernel to be patched and hacked) it's worked just fine. Under heavier load, other OS's (specifically FreeBSD, though NetBSD has occasionally made a strong showing as well) has been the OS of choice, simply because it has handled the file descriptor problem (along with a few other minor things) better than Linux.

    Personally, I'd be very interested in seeing some testing done with the new 2.2.x kernel series. I'd have to guess it would manage *much* better, as I believe most of the patches to increase the number of file descriptors has been integrated and tested and fixed, so it's working properly.

    To recap, running an IRC server with greater than 256 users is a Bad Idea, if you are using Linux and running a kernel prior to 2.2.x. Running an IRC server with up to around 1k users should be pretty good to go with a 2.2.x kernel. Above that, I don't feel qualified to say, as I've not tested it myself, yet.

    If anyone has done any real testing with ircd on Linux 2.2.x kernels, I'd be very interested in hearing the results. ;-)
  • IRC is a very specialized application that you dont see much of in the real world?

    tons of data being transmitted to thousands of participants whose lightweight clients connect to large, high traffic servers which provide a network on top of the Internet.

    how is this not a real-world system? isnt this a pretty common paradigm on the Internet?
  • If I saw this article written as a comment, it would have been moderated down as flame bait.
    Enough said.
  • by tgd ( 2822 ) on Tuesday September 21, 1999 @07:03AM (#1669126)
    I can't talk on the IRC issue, its been years since I've used IRC, but I seem to recall there being some sort of an issue with the networks themselves, not with the robustness of the platform. In the early days of Linux, I saw that a lot for Usenet, IRC, and other network tasks that "old school" administrators saw as being the realm of the "real" Unixes. Sometimes it'd be tough to get the feed in the first place, not to deal with it once you got it.

    The eBay comment, though, I thought was an interesting one. I could say with virtually 100% certainly that you could EASILY get a Linux-based system to perform more reliably than the current system, but that's not the fault of Solaris and itsn't a Solaris vs Linux issue, its simply a network application architecture issue.

    Dejanews has the right idea, boatloads of Linux systems with a good application architecture, and you'll never have any downtime.

    Admitted eBay has an interesting situation in that the nature of auctions where users can have bids automagically updated means running a lot of business logic on the database server, which can really move the bottleneck to the actual server software and the stored procedures, not the OS itself. My understanding is eBay runs Oracle, and Oracle has always struck me as being a real bitch to get good redundancy on and replicated servers that can cleanly fail-over.

    A site like e-bay could easily be reliably run on Linux systems using a well designed architecture though, with database servers (running Oracle on Linux) that are handling only portions of the site, and a lot of inexpensive front-end servers. Adding a tier in front of that, made up of http-accellerators, would make the system even more robust.

    Salon and dejanews both show that properly done, you can do high-profile with Linux. Certainly more so than you can with NT, particularly if you don't have millions to throw at hardware.

  • It seems Dvorak has missed the point yet again ... it's partly the lack of choice that has seen Microsoft in the hot seat with the DOJ. A lack of competition leads to market dominance, leads to a lack of dynamic efficiency and a loss of innovation, and eventually leads to products which stagnate. What reason is there to improve on your product if there's no-one to compete with? Might as well keep re-releasing the same product, just with different version numbers ....

    The fact that Linux has so many window managers is an advantage, rather than a limitation. It ensures competition, and ensures that development will continue to happen.

    I think the general opinion is that Dvorak writes not out of knowledge, but out of sensationalism. Look at the reaction he's got out of us ;)

    "No silicon heaven? But where would all the calculators go?!"
  • ,but for different reasons. Where does most of the development occur on Linux? Where there is an itch. How many people have an itch to run one of the Big3 IRC servers on Linux? Now compare that to the number of people and companies that use the multitudes of smaller systems that have been scratching for years to get rid of the fiberglass necktie that is Microsoft.

    Most of the coders for Linux do it in their free time on their own equipment. Most of those coders don't run superhuge servers of any type, can't afford such systems, and if they did they would most likely be happier to pay someone big bucks for a solution rather than hack it themselves. But a $199 system...aahh, now that is what I call hackable. Everyone can afford one, and go hacking on the kernel. Before long, it works better on the cheap system than anywhere else.

    Is anyone suprised by this? Development on Linux will take the path of least resistance. Lack of access is a big impediment; therefore, Linux will shine more and more on cheap systems. That doesn't imply that it will shine less on large ones, though. People there still itch, they'd just rather pay someone else to scratch for them.

  • Back in the days, ircd under Linux would fall over and die about when it hit 1000 simultaneous users (for the individual server, not the network).

    The main problem was the way you had to hack the kernel to allow ircd to open more than 256 file descriptors, and the fact that this hack didn't mesh perfectly with the rest of the kernel, because it was a hack.

    This problem was fixed, but not quickly enough to prevent most IRC networks from switching all servers and development to some form of BSD. Linux's failure on IRC also made the networks wary of trying it again.

    IRC servers on big networks use up enough resources that they're the only thing running on the machine, so switching OS's to suit the prejudices of the network you're applying to really isn't too much of a problem.

    Also, DALnet ircd development moved to FreeBSD, and the developers were loath to optimise it for Linux use, because since all the servers run FreeBSD anyway, what's the point?

    Charles Miller
  • Hey, the dvorak layout [] is not too bad. If your hands feel like pounding rocks after ten minutes at the keyboard, its something worth pursuing.

    Now Dvorak, the com^Hlumnist [], is another nut off the zdnet tree.
  • by Shane ( 3950 ) on Tuesday September 21, 1999 @07:09AM (#1669156) Homepage
    One thing you might find interesting is.. a good number of the people who develop freebsd also develop IRC. You might find it also interesting to know that the majority of these people spend most of their lives on IRC, and they all have global O:'s and most of them run HUB servers :)
  • Slash Dot Community-
    I read the Dvorak article, and some of the parts struck me as wrong. The overall impression was someone incredibly frustrated who wanted to take out his frustration on something. That's just what it seemed, by the tone and infliction of writing. Lets take a look at Dvorak's article and see what we can come up with.

    Ask Linux advocates exactly why no IRC (Internet Relay Chat) server runs Linux. One of my networking gurus said that he has never seen any Linux system on the IRC--probably the roughest networking environment on earth--hold up to the strain. In this instance, we're talking about the three major IRC networks: DALnet, Efnet, and Undernet. With tens of thousands of users logging on at once with numerous clients and multiple connections and with constant attacks from sources around the world trying to shut down the service, Linux can't cut it--period. Why is this never mentioned by the Linux proponents who flood the critics with e-mail anytime anyone says anything to disparage the OS?

    Well, information provided here isn't entirely true. I know of two servers down on Undernet running Linux. That alone right there proves that Linux can "cut it". As far as the Denial of Service attacks go, bring down all nonessential services. It's the first (and very useful) step in securing a server. For an IRC server, only run the IRCd, secure shell (ssh) and maybe BIND.

    Just hating Microsoft is not a good enough reason to promote Linux above everything else. When Linux fails in various sub segments of the marketplace, people always assume the problem is its open-source nature. "You can't get support." In fact, many companies support Linux in the same way Sun supports Solaris. Linux often fails, simply because it isn't robust enough. Contractors know this but fear the wrath of the Linux community and blame support. It's a phony excuse. This BS factor is what bothers me. The inadequate-support explanation creeps in for no other reason than to keep the lunatic fringe of the Linux movement from clogging the e-mail system with complaints. The fact is, Linux has yet to prove itself at the top of the food chain, and until it can run eBay, for example, it will remain the "in-between" OS.

    Dvorak does make a valid point with the first sentence. Hating Microsoft should not be the only reason one promotes Linux. Have reasons. I personally think that my uptime is a sufficient reason (at 65 days right now), which is also closely linked to system stability. I've run NT servers before, and I'm running one now. I just got done installing one about a week ago, and besides the persistent reboots from installing software, it is doing okay so far. It did crash numerous times during installing (It hated the ATI Mach 64 card, I had to replace it). We'll see how the new one goes, the other NT box I admin I have to reboot weekly. This isn't anti-Microsoft propaganda (well, maybe it could be viewed as such) however it is the truth. But back to the paragraph, the question on support is valid and not at the same time. I personally have never had to go to a support company, but oftentimes my problems with Linux have been misconfiguration issues that I had to resolve. As far as the eBay comment goes, well, I wonder what the eBay system runs for a backend. I would bet not NT, I would bet probably SunOS or Solaris on a nice Sun box.

    The low end is where Linux should gravitate. By this I don't mean Pentiums running in offices. I mean on AMD chips in sub-$200 computers with small amounts of memory and $50 hard disks. Under a normal load Linux is quite remarkable. It's small, fast, and stable. It's quite amazing actually. But it's still Unix and used primarily as a command line OS. There are numerous GUI shells for the thing, and there's no reason the Linux community can't standardize one and stick with it (except that there seems to be a rule against even imagining such a concept as agreement).

    Okay, another point which hits and misses entirely. Yes, Linux does perform exceptionally well on lower end equipment and older equipment. The performance is much better than with Microsoft software. Does that mean we jail it there? No. It almost seems (I may be out on a limb here) that Dvorak thinks since the MS software has been around for so long and has the user base (mainly due to lack of competition) the software has gained some right to always be the only thing running on higher end (and in Dvorak's case Intel) machines. Well, unfortunately enough (for Dvorak) the AMD chips perform just as well if not better (performance depends on environment and application) than the Intel chips, and MS software has no innate right to be run on higher end machines. Software has to earn the right to run on machines, it is not written (or "born") with that right. It is not like the Royal Family of England. As the processor levels go up, so does the performance on Linux machines. Of course, the same is true on MSOS machines, but because it is true there does not make it untrue elsewhere. The next point Dvorak makes is that Linux is still UNIX - Yes, Linux is a UNIX variant. True enough. The claim that it's primarily a CLI (command line interface) is another question. True enough, most remote administration is done on a CLI through telnet or SSH. Local administration and use is up to the operator. I know of many Linux users (mainly friends newer to Linux) that live and breath in the KDE and Enlight/Gnome window managers (wm), and I know NT server administrators that spend 90% of their time in DOS boxes. Dvorak's next claim that the increased usage of CLI is because of a lack of standard or common WM was incredibly hilarious and once again false. Choice is not an evil thing. Lack of choice can be considered to be, but is not necessarily the case. Please, Mr. Dvorak, lets not eliminate choice and free will here.

    Now that you can combine a free OS with any number of cheap Linux-oriented office suites, it's time for a reemergence of VICs (very inexpensive computers). Microworkz has a $199 machine called the iToaster, which uses the kernel of the BeOS (the true competitor to Linux) to keep prices low. This trend will emerge within the next 12 months and will take the world by storm. Intel and the chipmakers know this. In case you haven't noticed, their strategies are going toward the low end in a near panic.

    The OS running on a "VIC" is not the only contributing factor to the cost. They are also running with inferior parts. A lower quality hard drive (yes it makes a difference in performance AND reliability) and a lower quality processor (example: Cyrix) can contribute to the overall cost, reliability and stability of a "VIC". The cost, reliability and stability all go down. I used to work at a Best Buy store (as a technician who backed up to the sales floor), and I can reliably say that people looking for "Bargain Basement" (our term for VICs) computers were the first time computer users wanting to get a computer and see what it is like, without investing a lot of capital. After their first one, they would start looking at the higher end machines. Occasionally they would bring their machines back within the allotted return period to pick up one of the higher end machine. The reason given, most of the time was either reliability or performance. For our Best Buy store (at the time) the higher end machines were Hewlett Packard, followed by IBM, then Compaq, Acer, and then Packard Bell/NEC. The Acer, PB/NEC, and the eMachines were considered the "Bargain Basement" of the bunch - often run between $600 and $900. Cheaper should not be considered better. On the next point, the true competitor of Linux: How is Linux truly competing with anything. From the current social context of the word "competition" when related to business or computers typically comes down to sales and profits. Last time I checked, we could still get Linux for free from sites like on the net. Now I suppose if you wanted to compare user bases, then all OSes are competing with all other OSes. Every OS has its various strengths and weaknesses, which will be discusses momentarily.

    However, I would like to jump back to the IRC server question for a minute. I have played with the IRC source from the Undernet and DALnet servers before, on Linux machines with kernels ranging back from 2.0.32 up to 2.2.9. I will admit that IRC server performance under Linux (at least the 2.0.xx series) was not the greatest. Also, server capacity was limited. Under the 2.0.xx series kernels, by default you could only have 256 file descriptors per process, which limited the number of incoming server connections an individual server could handle. This is limiting in server performance for the higher traffic and load servers, such as DALnet. Now, I did cruise on through the DALnet and Undernet sites, and I grabbed a bit of information to share. This first quote comes from the DALnet site.

    Linux has been found not to be workable with DALnet's traffic load. All DALnet servers running Linux have been switched to FreeBSD. Therefore, if you are currently running Linux on the server for which you are applying, we suggest you switch to FreeBSD as soon as possible, preferably before sending in your application. Servers running pre-2.0.31 Linux kernels cannot be linked to DALnet at this time.

    Well, yes, under the 2.0.xx kernels, it was not the greatest of things to run the IRC daemon under Linux. Performance from identical machines running BSD was much better, and so was the reason to switch over. Now, I have not yet run with a high traffic load on a server running the 2.2.xx series kernels yet, I was going to start to play with traffic load in a few days. I know that under the 2.2.xx kernels there are 1024 file descriptors per process, and that does help with the ability to host clients. A lot.

    The next quotes up are from the Undernet site:

    Stable UNIX host available for 24 hrs a day with an average of a week's uptime. The kernel should also be able to handle a minimum of 1024+ open fd's for clients, and the server must be configured to allow that many client connections

    The operating system must support a minimum of 1024 fd's per process.

    Please also include information about the hardware, such as the amount of memory, and what other services the machine runs or will (not) run, such as HTTP/news/mail servers, and what other restrictions the machine has, such as a limit of 256 fd's per process. Output from the uname -a, limit/ulimit -a, and dmesg commands are required. Minimum Hardware: Pentium 266Mhz (or comparable if sparc, SGI, etc...), and 128MB of RAM (256MB recommended).

    The only things the Undernet network is looking for is a Unix based host (Linux applies) and a minimum of 1024 file descriptors per process. Now, it is possible to adjust the number of file descriptors on the 2.0.xx series of kernels, I just never went that far. I will discuss this more in a minute. I have one last quote from the Undernet site, and here it is:

    The only ircd acceptable for use on the Undernet is the Undernet ircu2.10.xx, which is the current release software. Although a windows NT port of the software has been written, windows NT servers will NOT be permitted to link to the live network. The NT-based server version was written as a proof of concept experiment, and not meant for production servers on the Undernet.

    It looks like Windows NT is not considered stable or reliable enough (those are the reasons given for the decline elsewhere in the site) to connect to the live network. IRC servers run very well under UNIX, but performance differs from flavor to flavor.

    I have a friend who administrates the box that a higher capacity web server runs on. He has upwards of 40 hits a second, and transfers 50 Megs of HTML and images in about 10 minutes of uptime to requests all over the world. He was running Linux, back when the 2.2.0 kernel hadn't been released yet. His server's performance was well below acceptable, and Apache was having uptime problems. After anywhere between 5 minutes and 10 hours, the server would "freeze up" and not operate anymore. It would have to be restarted (Apache, not the system). He had a UNIX programmer friend (mutual) come over and heavily mod his kernel. (The friend is an accomplished Linux and BSD programmer). He upped the file descriptors per process to 1024, and performed a large number of tweaks and optimizations to the Linux kernel. The server still wasn't performing as well as it should have been. I went in and redid his apache configuration file. We went through and optimized the entire machine. We still couldn't get reliable performance from it. We ended up switching the machine over to FreeBSD, where it also runs the Apache web server. The server is now performing acceptably, and has not had an uptime (Apache, not system) problem since. For reference, the server had 256 Megs of RAM, about 6 gigs in a SCSI U2W chain, and as I recall, a 300 or 250 MHz processor.

    The point of the long, rambling paragraph being that Linux is not the only answer. I am a Linux advocate, and I will promote its use wherever applicable, however, it is not the only answer. Sometimes BSD performance is better. Sometimes (but rarely, in my experience - mostly limited to people who didn't want to go a different was) NT is the answer. Sometimes Solaris on a SPARC is the answer. We have to go with what works for the situation. With the continuing Linux development, for most things, Linux works just great! Fabulous even.

    But wrapping things up since I've managed to confuse myself with the direction I took with this: Some of the points Dvorak made were on base. A lot were not. However, there are always two sides to an issue, and oftentimes more then that. We should look at them all, and consider the applications for it. Just because Linux may not be the greatest at hosting an IRC server doesn't make it any less of an Operating System. Windows NT can't do it reliably either. If you want to go out, and rewrite an IRC server and make it best for Linux, go right ahead. Do so with my blessing. If you don't, that's perfectly fine as well. We can't afford to get caught up in squabbles over which UNIX derivative is the best - all UNIX is good UNIX. Boost it where we can, support it where we can, because we of the Linux community are trying to show people another way. Its not the only way, but it is another way.

    -Captain Keen

  • by Hollins ( 83264 ) on Tuesday September 21, 1999 @07:10AM (#1669168) Homepage
    • Dvorak seems to dismiss Linux/Apache as a server contender because he fees it cannot currently handle super high-load jobs, and is thus only well-suited for medium instensity servers. These 'in-between' servers, as he calls them, are the workhorse of the internet and represent most of the servers out there. I don't see how he can contend that filling this niche is not a victory for Linux. He doesn't attempt to address the dizzyingly fast evolution of this OS, or admit that it very well may be handling these high-end applications in the future. I am very happy to see Linux used for all but the most demanding server tasks, which are still relegated to Unix. This is a validation of free software, not an arguement against it.
    • He thinks Linux will own the desktop because MS will sooner abandon its market share than price Windows competitively for $200 machines. There is no way MS will give up this market. They will either price Windows down, or come out with a cheap version in between what are now Win98 and WinCE to be used in iToaster-type machines. Expect it to be Win98 with a few key dlls removed (much as NT workstation is NT server sans a few dlls). I expect they'd price it at $10-$20 OEM. It is absurd to argue that MS is going to roll over and give up Windows on the home desktop without more justification than Dvorak gives.
    • UnderNet may not run Linux, but my web hosting service (Hurricane Electric) does. They host a lot of relatively high-traffic stuff. I haven't noticed any down time in the two years I've been using them. They let me telnet in, run custom scripts on the server, have PhP and MySQL, and charge less than $10 a month. They're also really fast. I haven't seen anyone able to provide this price/service combination on an NT or Unix system.
    Dvoraks columns seem to be hit or miss, sometimes he's dead-on insightful, but others I think he goes too far to deliver a viewpoint outside the mainstream. I fear this is more a case of the latter.
  • ha, we've heard THAT sentiment before. Personally, I feel that the typical Wintel box of today, the latest and greatest, is a horribly backwards machine compared with what could be and will be. Ford may have mass produced and made an affordable car, but they have evolved tremendously since then - and the typical wintel box is still at a very primitive stage, w/ much room for improvement. It may take a revolutionary discontinuity instead of a smooth progression, but it will happen.


  • by Anonymous Coward on Tuesday September 21, 1999 @07:12AM (#1669186)
    I can remain silent no longer. It is time that the manufacturers of internal combustion engines accept the fact that no IC engine can drive the major steamships of today. These engines are fine for emergency power on your weekend sailboat, but it takes coal, lots of coal, to turn water into steam to move the Big Ships across the Big Seas!

    It's time that IC supporters recognize the fact that there are exactly two modes of transportation in the world today, million tonne displacement steam ships and single passenger motorscooters (notwithstanding the fact I just mentioned a third, pleasure watercraft), and that since they can't drive steam ships (today or any day, so say I), then they should focus on the only other market niche in existence!

    (Silly? No sillier than Dvorak ranting and raving about Linux's perceived (or simply alleged) failure to perform well in certain extreme causes while ignoring the 800-pound elephant turd in his eye that is NT servers pressed into duty to replace "doesn't require a 20-MB GUI front end, so it must be bad" Unix server.)
  • by bombtrack- ( 64192 ) on Tuesday September 21, 1999 @07:12AM (#1669187) Homepage
    I have coded and worked with IRCD. wIRCD, windows varients are rather unstable. And no, not all ircd is devloped for BSD. Most of the people that are responsible for Hybrid, the efnet ircd, run boxes (with linux) that run ircd fine. Elite IRCD runs better from linux than bsd. I guess it really depends on what version of irc you want to run. I'd be happy to answer any IRC related questions you have.
  • While it does have a slightly bad tone to it, the article makes a few good points:

    1) People gravitate to linux because it's the underdog
    2) Linux is not a top notch server
    3) Linux is ideal for embedded and semi-embedded computers

    Linux outperforms many other systems in serving in many situations, but it's not 'quite there yet'. (I know, it will be. Real Soon Now (TM))

    I think a main reason for him to write such an article is to counter all the hype surrounding Linux. It is absolutely true that there are many people and businesses out there that are applying Linux to their problems because of the hype surrounding it. Many of them are failing, perhaps because Linux isn't a good solution to their problem, partially because they aren't putting forth the effort to make it work, etc. But mostly because they are being told from every angle that Linux is the new kid on the block that fits every hole possible. It's not there yet.


    "You stay here, while we go get help..."
  • Personally, as far as ZDNet selling out to Microsoft, I think if it's being done (and I am of the opinion that it _is_ being done), it's not a company-wide policy, but rather a few so-called journalists doing whatever the Redmond PR department tells them to. I've seen some positive press for Linux from some people at ZDNet, notably the SmartReseller bunch.

    I want to let you know I agree with this. ZD has a Linux section on their site, which seems quite positive. Also, I read Smart Reseller myself, and get the hardcopy magazine, and can't think off-hand of any time I've been really displeased with anything in it. Those in ZD who consistantly knock themselves over their behinds with illogical arguments against Linux are just a loud-mouthed minority.

    More controversy equals more page served equals more banner ads equals more dollars. That's probably the sole reasoning behind his columns.

    Yep, they probably made a killing of that article. The shame of online articles with per click banner ads. It's not important to have real reporting and journalism anymore. Crap which gets people to the site is more worthwhile then a true article that no one notices.

    It's a shame...

  • by Shane ( 3950 ) on Tuesday September 21, 1999 @07:20AM (#1669205) Homepage
    Has anyone else noticed that it is becoming more common to hear a *bsd slanted article where the author talks about Linux in the same way we talk about Windows?.

    Something is wrong with this.. its not how its supposed to be. We are supposed to get along with our *bsd brothers and sisters. We are not against them.. we are with them.. for the good of software everywhere :).

    Every Free OS installed is a wonderful thing for us all. every installation of Linux at the expense of Freebsd is a BAD THING.. every Freebsd install at the expense of Linux is a BAD THING.

    Picking the best OS for the job is one thing. Spreading FUD with no facts to back it up is another. If linux really can't cut it as a Efnet IRC server, where are the facts? HUB admins have not allowed someone to link a major linux server so how do they know if it can not perform?
  • Keep in mind that, to the best of my knowledge, Dvorak has made his living commenting on the PC world, and probably is not a Linux User. As such he is biased. Even before I started using Linux, i would read his article in my dad's PC Mag and wonder what he was smoking on occasion.

    It seems to me that he thinks Linux is a finished product, a permanent freeze on the source tree. But in reality Linux is constantly evolving.

    He is correct that Linux is not ready for the "big time", where Big Time is high end high SMP servers. But we know this. There was an article on /. last month comparing NT and Linux to the high end servers. One example I remember is about Linux not being able to hot swap drives. (am I wrong on this?), and the SMP support is still in development. I believe that the high end servers Dvorak refers to have 16+ processors, the area where Solaris and its kin rule. And then there's that whoel SGI filesystem deal we've all been raving about... Evolution happens a little bit at a time.

    Big systems on big loads? Excuse me. There is this MIT (I forget the name) company that has a patent for a web load balancing algorithm. There was a /. about it last month or so. They have hundreds of servers spread accross the world. Amoung thier clients are ESPN and I believe CNN. That is a major load. What to they run? Linux. Hacked to their specs, but still Linux. And old kernel rev too.

    What about Beowulf and other clustering software? What about all the research centers running Beowulf clusters? Like NASA?

    Linux often fails, simply because it isn't robust enough isn't this in contradiction to all the other beneficial press about Linux? It may not be the Honda of OS's. Buy if you tune your '69 mustang it'll purr like a kitten.

    There are numerous GUI shells for the thing, and there's no reason the Linux community can't standardize one and stick with it

    So? It isn't about standardization. There are so many shells because people compete against each other for the best solution. Which gives the best code. It's also about choice.

    He is right about the VIC's though. Linux could be perfect for them, and as a thin client. But it could still be a powerful server. A server running the VIC's and thin clients.

    This article, although incorrect on many points is good. Why? It pisses people off, who then go start developing on their own.

  • With Linux gaining acceptance for very large scale applications (Weather supercomputer and other applications of Linux clusters to computationally intensive problems, mostly for cost reasons (commodity hardware vs. custom)), it seems that some people think that Linux is up to or can be tuned to be up to the job of handling some of the largest applications around...

    Things like IRC and Oracle stability are an issue of applications interacting with the OS, and perhaps these apps need patching and tuning as much as the Linux Kernel does. Many of these issues are issues of people comparing entire systems (including applications, and system admins) without taking into account the system variables (is the application ported? was it tuned for that OS? are the system admins of the two systems being compared of comparable skill?)

    Clearly Linux has its weaknesses, but when analyzing it is much more useful to do a straight comparison than the touchy-feely anecdotal kinds of comparisons that people tend to make... and to know what you're comparing. To say "Linux works poorly with IRC server software" is not to say "Linux is bad for all large scale server apps"...
  • Ok, I read the article and the first thing I thought was "wow, this guy is full of whatever that stuff is that they make hotdogs out of"...

    But then I read it a second time, and realized he's half right.

    First, the instant retaliation of pointing out NT's faults when someone mentions a fault of linux must be stopped. Linux isn't always the best, as he said...but it's often better than most.

    I personally don't think Linux makes a great desktop, because it wasn't made to be one (though it could be with work). For a desktop I prefer BeOS, but for a server, I've yet to have linux go down on me. But then neither have any of the FreeBSD servers I've worked on, nor any BSDi servers. The NT machines I work with at work, well, lets not mention those.

    I do however agree that where linux shines is medium-strain servers...'why?' you ask, well, it's simple, because there are more medium-strain servers out there than huge servers, so there are more medium-strained-system-admins to use linux, and fix what needs to be.

    Linux will be able to handle a highly-strained server when it needs to be, because when it needs to be someone will make it work.

    Hopefully I didn't just TOTALLY repeat what everyone else commented on...I haven't read them all yet...

  • by Christopher B. Brown ( 1267 ) <> on Tuesday September 21, 1999 @07:44AM (#1669241) Homepage
    The reasons why Linux is not used on IRC are not particularly related to Dvorak's discussion of how "Linux is apparently not suited to IRC."

    For that reason, it is reasonably fair to say that the Dvorak essay is significantly flawed.

    It is, after all, not particularly informative when Dvorak comments that Linux isn't used on the big IRC servers when he makes no comment on what is actually used for the purpose, or why.

    The only implication that the gentle reader could reasonably guess, from the vendors mentioned in the article, is that the big IRC servers are all running on NT.

    Is it therefore a "Pro-MSFT, Anti-Linux" piece? Probably not.

    It's certainly not a "Well-written, well-argued, well-defended" pice.

  • Win 95 *does* run fine under low-end systems. I've run it on a 66Mhz 486 with 16MB of RAM, which is pretty low-end these days. It was plenty unhappy with only 8 meg, though, resulting in a lot of hard drive thrashing. It didn't, however, crash... just slow as molasses.

    Linux, in my experience, does run better on the lowest of the low (16Mhz 386sx... 6MB of RAM, had a BogoMIPs score of 8 or something like that). Granted, however, that was running without a GUI. I don't think Win95 would run well on that system, but then it can't really do without a GUI, unless you're just talking about running DOS (which is no match power-wise for Linux).

    Generally, it's not the OS that kills your system, it's the apps. Linux runs OK on my 33Mhz 486sx laptop. Finding things to run on it, though, is a challenge. Netscape will thrash the drive for about a minute before it comes up. StarOffice probably wouldn't run at all. Older versions of Siag office ran OK (haven't tried it recently) and at least let me use it for light-duty word processing.

    What was probably crashing your machine was some sort of misconfiguration, not a lack of processing power or memory. Lack of either just means the thing is slower, not crash-prone. Now, the fact that Windows got into a crash-prone state is something you can rightly hold against it.
  • by Bruce Perens ( 3872 ) <> on Tuesday September 21, 1999 @07:30AM (#1669246) Homepage Journal
    For something like 20 years now, I've been reading articles about how silly Mr. Qwerty is. Other authors wonder, in print, how magazines keep paying him and why anyone still reads him. And yet, Mr. Qwerty writes on. Perhaps just because he's the author everybody loves to hate.

    Linux technically inadequate for IRC? Rubbish. No doubt 50 people will go out and prove that this week just to make Mr. Qwerty look bad.

    Much of the success of Linux, by the way, might be attributed to the fact that Linux folks are busy writing software instead of hanging out on IRC. I doubt the fact that we're not the main host of IRC servers has much to do with Linux' current technical limitations.

    Bruce Perens

  • by davidw1 ( 29211 ) on Tuesday September 21, 1999 @07:34AM (#1669251)
    I think Dvorak was just wanting to get more fame by putting this crap on zdnet. * Linux 2.1.xx * Linux 2.1.xx

    I'm sure there are more.

    Dvorak: how about checking stuff before you spew it out?
  • At least FreeBSD, OpenBSD, and NetBSD can compete on their merits, instead of by buying/crushing/smearing the competition.

    I love Linux, and I'm setting up my firewall/NAT box at home on FreeBSD. Why? Because less hackers are familiar with it, I have loads of good tuning information on it, and it works great for most of our production servers at work. Not that we don't have about 100 of each FreeBSD and Linux. And not that I don't run Linux on my laptop. I just feel better about having a firewall at home that is built on an "old-school" type of distro.

    Competition is the inevitable result of scarce resources. Dvorak is being an idiot; the BSDers aren't. They have a good NOS and it shows.

  • by Lord Kano ( 13027 ) on Tuesday September 21, 1999 @08:08AM (#1669259) Homepage Journal
    This is a completely valid response. This is Dvorak's evidence that linux isn't ready for "The Big Time". Well if the M$ flagship can't do it either, why is he bitching about linux? Why isn't he making the observation about both?

    When unfair criticism is made we should point it out. You will never win if you allow the terms of victory to be defined by your opposition.

  • by bmetzler ( 12546 ) <bmetzler@ l i> on Tuesday September 21, 1999 @08:16AM (#1669278) Homepage Journal
    We do ourselves a disfavour by assuming that anyone criticising Linux has ulterior motives -- promoting NT, for instance, as many people seem to think.

    I the most part I agree with your statement. But remember, although he didn't specifically state NT in his article, he did say that Linux at least needs to be good enough to power eBay, which is currently "powered" by NT. I assume he was talking about the front-end.

    Not that Linux *can't* power eBay. We don't know that, we just know that eBay has decided to use NT. Did they ever do some field trials and determine that Linux couldn't handle the load? Or did they just choose NT because that's what you were supposed to use on the servers?

    The whole point of his article was that because Linux couldn't do everything, it shouldn't do anything. (except power low-end black boxes) But that's wrong. When I am evaluating what to run an e-mail server on to handle 120 users, one of my criteria is *not* whether the OS can handle 10,000 IRC users.

    There are many OS's that all have different strengths. So let's use each OS to it's fullest strength, and not restrict our OS choice to one OS that supposedly can do "everything". Use FreeBSD for your IRC servers, use Solaris for Database, use linux for web servers. No problem with that, is there? Dvorak seems to imply that there is

    I don't know about you, but I don't advocate using Linux in any area where there is a better solution. Does anyone? Oh, except drooling, adolescent kiddies on /. that don't really have any position of influence.

  • Truthfully, I think he was taunting the slashdot crowd to come on over and post your flames. The more feedbacks they get the higher their ratings. Just like the Neilsen ratings. Show alot of swimsuits, flame Linux -- same thing.

    Exact-o-mundo! ZDNN is nothing but a big banner serving machine. Talkback is an interesting mechanism - every talkback is presented on its own page with a full complement of ads. If 50 get posted on any given "story", and 100,000 readers view each one, ZD's "eyeball count" leaps by 5 million, driving up their banner rates and swelling their number of served impressions.

    Now the best part is - the readers provide the content for nothing. The no-cost content is packaged into pages and served up generating all kinds of revenue. If you thought they were just being nice and letting you speak up, well, think again.

    Bite My Ziff, Davis!

    "Cyberspace scared me so bad I downloaded in my pants." --- Buddy Jellison

  • I'd imagine the busiest site running Linux would be, or one the myriad other enormous sites that run Linux.

    As for your pointed stab at the poor guy's NIC, maybe you'd be willing to donate the $20? 'Till then, it'll work fine in Linux, hey.

    - A.P.

    "One World, one Web, one Program" - Microsoft promotional ad

The trouble with the rat-race is that even if you win, you're still a rat. -- Lily Tomlin