Linux May Soon Lose Support For the DECnet Protocol (theregister.com) 69
Microsoft software engineer Stephen Hemminger has proposed removing the DECnet protocol handling code from the Linux kernel. The Register reports: The timing is ironic, as this comes just two weeks after VMS Software Inc announced that OpenVMS 9.2 was really ready this time... That announcement, of course, came some months after the first time it announced [PDF] version 9.2 [...]. The last maintainer of the DECnet code was Red Hat's Christine Caulfield, who flagged the code as orphaned in 2010. The change is unlikely to vastly inconvenience many people: VMS is the last even slightly mainstream OS that used DECnet, and VMS has supported TCP/IP for a long time. Indeed, for decades, the oldest email in this reporter's "sent" folder was a 1993 enquiry about the freeware CMUIP stack for VMS.
One of the easier ways to bootstrap VMS on an elderly VAX these days is to install it on the SimH VAX hardware simulator, and then net-boot the real VAX from the simulated one. Anyone keen enough to do that will be competent to run an older version of Linux just for the purpose. Although their existence is rapidly being forgotten today, TCP/IP is not the only network protocol around, and as late as the mid-1990s it wasn't even the dominant one. The Linux kernel used to support multiple network protocols, but they are disappearing fast. [...] For a long time, DECnet was a significant network protocol. DEC supplied a client stack called PathWorks to let DOS, Windows and Mac clients connect to VAX servers, not only for file and print, but also terminal connections and X.11. Whole worldwide WANs ran over DECnet, and as a teenage student, your correspondent enjoyed exploring them.
One of the easier ways to bootstrap VMS on an elderly VAX these days is to install it on the SimH VAX hardware simulator, and then net-boot the real VAX from the simulated one. Anyone keen enough to do that will be competent to run an older version of Linux just for the purpose. Although their existence is rapidly being forgotten today, TCP/IP is not the only network protocol around, and as late as the mid-1990s it wasn't even the dominant one. The Linux kernel used to support multiple network protocols, but they are disappearing fast. [...] For a long time, DECnet was a significant network protocol. DEC supplied a client stack called PathWorks to let DOS, Windows and Mac clients connect to VAX servers, not only for file and print, but also terminal connections and X.11. Whole worldwide WANs ran over DECnet, and as a teenage student, your correspondent enjoyed exploring them.
Microsoft proposes? (Score:1)
Re:Microsoft proposes? (Score:5, Informative)
That's because now there's an x86 version of OpenVMS [vmssoftware.com] available.
Re: Microsoft proposes? (Score:5, Informative)
Apply for OpenVMS V9.2 Evaluation License
To help us qualify your application for this evaluation release kit, please fill out the following survey. Please be advised that community (hobbyist) licenses are not currently available, evaluation licenses are provided to commercial customers only.
So I guess I can't get rid of my microVAX, at least. It yet...
Re: (Score:2)
So I guess I can't get rid of my microVAX, at least. It yet...
Do yourself a favor and use an emulator.
Re: (Score:2)
The emulators boot too fast. They also don't provide the sort of radiant heat in winter that a microVAX does,
Re: Microsoft proposes? (Score:2)
Seriously, you want to argue in support of keeping DECnet support in Linux simply because it was suggested by a Microsoft employee - a full 12 years after last developer supporting the Linux DECnet code called it "obsolete"?
This has literally nothing to do with Microsoft -at all - but you can't help but be triggered by the mere mention of their name?
Re: (Score:2)
And why they have to be mentioned? They are opposite realm of computer culture. Far away.
Re: Microsoft proposes? (Score:4, Informative)
If you are using Windows shares today or use a Windows network printer, you are using a TCP/IP re-implementation of SMB based on TCP instead of DECnet. So, Microsoft Networking was the last big use case for DECnet on Linux besides bootstrapping VMS on old DEC hardware.
Re: Microsoft proposes? (Score:5, Informative)
If you are using Windows shares today or use a Windows network printer, you are using a TCP/IP re-implementation of SMB based on TCP instead of DECnet. So, Microsoft Networking was the last big use case for DECnet on Linux besides bootstrapping VMS on old DEC hardware.
No. To my knowledge there never was a SMB implementation over DECnet. PATHworks contained various things: the client included DECnet for Windows; the server included an SMB file server which ran over NETBEUI. NETBEUI was developed by Microsoft and IBM back in the day and was a simple non-routable protocol.
Reverse engineering how the PATHworks server worked led to the birth of SAMBA but no DECnet was involved.
Re: (Score:2)
How could you fantasize like this? SMB is not DEC, why couldn't you tell this much.
Re: (Score:2)
Re: (Score:3)
Seriously, you want to argue in support of keeping DECnet support in Linux simply because it was suggested by a Microsoft employee - a full 12 years after last developer supporting the Linux DECnet code called it "obsolete"?
Orphaned, not obsolete. Orphaned means no developer is still working on the kernel code. Obsolete means almost no users are using/installing the kernel module. To be fair, it is probably both, but still.... :-D
Re: (Score:2)
Must be bad, deny.
Probably, if their proposal included supporting NetBEUI in the place of DECnet.
Re: (Score:3)
Same person who got rid of IPX and SPX.
Re: (Score:2)
I want my banyan vines and xns back.
csb: I still have my cisco employee 'list of all protocols' shirt from the early 90's. something like 'everyone talks about it' and lists of protocols like x.25, vines, appletalk, decnet, etc etc.
pretty famous shirt from back then. every newhire got that when they joined.
it really was something: the list of protos they supported included everything that mattered. it was a nightmare carrying all that in the base code.
Re: Microsoft proposes? (Score:2)
Stephen has been a Linux maintainer and contributor far longer than he's been at Microsoft. There's a good reason he's also known as the Network Plumber in the kernel's network stack.
Re: (Score:1)
Sometimes Microsoft is accidentally right.
If the usage of a sub-system drops below a certain percent, it should be removed and turned into a pluggin.
In college (Score:1)
I found an 11 780. RPI site facilities hooked up the electrical outlet in our club room incorrectly and swapped neutral with a phase or got the phase rotation wrong, I cannot remember. When turned on a lightning bolt shot through all the system planars and I knew we just lost a working piece of history.
Re:In college (Score:4, Interesting)
Re: In college (Score:1)
Nope! My time was about a decade later.
Ah, the memories... (Score:5, Funny)
We had one of these worldwide DECnet WAN infrastructures around 1996 or so. Bloody system came down every week with lots of meetings, RCAs, with everybody and their dog etc.
Then one week it didn't. And the next week neither. And nothing in the third week.
Then we were called into a meeting to explain why the DECnet trafic didn't go through their DECnet nodes anymore but all went through our Cisco routers, despite that they explicitly had told us not to do it.
Re: (Score:2)
Also remember, a big computer worm occured on SPANnet (space physics analysis network). The "Father Christmas Worm" because it wsa just before Christmas in 1988, and also less than two months after the more famous Morris Worm. So the SPANnet worm also made it into the mainstream news briefly.
The thing is, there was so little network security in VMS. Neither DECnet nor IP had security in the protocols themselves, they relied upon the host computer to do the security as to who is allowed to do what over a
IPv6 (Score:5, Funny)
If it takes this long to migrate from DECNet, IPv4 will be around for a long time yet ... :)
Re: (Score:2)
If you make the next "killer app" that will require IPv6 for full functionality and state "contact your ISP for IPv6 to get full functionality" then you might get things moving.
Re: IPv6 (Score:2)
Naw, companies would rather proxy everything through "the cloud" so they can make even more money harvesting data and forcing obsolescence when they shut down the cloud service it depends upon 20 months after you buy it than, well, make their device directly-accessible via IPv6 without involving them as mandatory middlemen.
Re: (Score:3)
The killer app for IPv6 is basically any use case that is not content consumption. Things that most ISPs don't like or explicitly forbid. They'd love to put everyone on a CGNAT and forget about IPv6.
Re: (Score:2)
Our devices are all IPv6. However the customers usually connect to them via IPv4 because that's the practical way to get a secure tunnel via a public internet provider. Overall, IPv6 is simpler even if it does mean tables get larger. In a world where people aren't shy about using 32-byte UUIDs to identify entities on a single computer, why they're upset about having to use 16 bytes for a world wide address.
Re: (Score:2)
Re: (Score:2)
The original DECnet supported only 2 nodes. Phase 2 allowed 32 nodes! Phase 3 (1980) allowed 255 nodes, and eventually phase 4 and 5 allowed a whopping 64449 nodes! Of course, these weren't internets, the use was to support communications within a company or school, or with a limited set of like minded entities. It was expected you knew all the nodes, much like the early ARPAnet days where you'd have each machine with a unique name, worldwide (kremvax). In that sense, 64,449 nodes is more than anyone w
Linux? DECnet? Say what? (Score:5, Interesting)
Back in the day I worked for DEC, a company with neat technology and total buttheads deciding what to do with it. I worked with DECnet (mainly via PathWorks) and knew it well. Making it run in 640k and having enough memory left over to do anything else was a challenge.
I didn't know Linux even supported DECnet. Shows how much attention I've paid to it since then.
...laura
Re:Linux? DECnet? Say what? (Score:4, Informative)
In the late 90s/early 2000s I used Linux to bridge all sorts of networks. You could have a file server connected to Novell, DECnet, SMB, AppleTalk, and serve the same files over NFS on TCP/IP.
It was probably how Linux took over the server rooms. It was the glue that could connect to anything.
Re: (Score:2)
I worked at DEC (at the mill, MLO) for a few years, and in the boston area, overall, for over 5. DEC was the major force in the area and everyone knew who they were. DEC stuff was great. customers liked it and it was very reliable. we had actual designers who thought things out.
I like telling the story about the painted line colors on the walls in the mill, the main corp HQ in maynard, mass, back when they were still around. the mill was over 100 years old and buildings were added over the years and br
Re:Linux? DECnet? Say what? (Score:4, Funny)
I do lots of dead bug electronic construction, which includes using high-value resistors as standoffs when needed. My current supply is a bag of 22 megohm resistors I got cheap at a surplus store. The old-timers look and have to count "red red BLUE?!"
...laura
Re: (Score:2)
Re: (Score:2)
Re: Flamebait article (Score:2)
You dun did it now...
Just for laughs I searched for "SMB alternative" and my browser had a huge message at the top, "There is no alternative to SaMBa"... whelp
Re: (Score:2)
They forgot about AFP/Appletalk... as always.
Re: (Score:3)
Hi there. I wrote the article. Did you actually read it?
Because I specifically mentioned AppleTalk, as well as IPX/SPX.
Re: Flamebait article (Score:5, Funny)
Hi there. I wrote the article. Did you actually read it?
Who has time for that when shit posting is involved.
Re: (Score:2)
At least RingTFS (then mouth off, but before posting) at least skimming TFA used to be the baseline. Even that is strongly contra-indicated in the modern slashdot world of clickbait headlines and random copy/pastes that really aren't summaries at all.
Re: (Score:3)
>Did you actually read it?
What part of "slashdot" don't you understand? :)
you must be new here.
[duck]
hawk
Re: (Score:2)
[duck]
hawk
Grouse
Re: (Score:2)
get off my lawn!
Re: (Score:2)
Hi there. I wrote the article. Did you actually read it?
Because I specifically mentioned AppleTalk, as well as IPX/SPX.
Hi, how do you do?
I actually read a part of it, thanks. Can't remember if you mentioned XNS, though :)
Note that when I wrote "they forgot" I meant the people who think "there's no SMB alternative", rather than the article or its authors :)
Re: (Score:2)
Hi. :-)
A part?! It's only about 500 words long. It's a short! :-(
No I do not mention XNS. Has Linux _ever_ supported XNS? Not that I'm aware of.
Ah, DECnet (Score:3)
Lovingly referred to as DRECKnet by many of its users.
Re: Why does protocol software need "maintenance"? (Score:1)
Platform changes are one obvious reason. Compiler changes are another.
Example from history: switching from 32-bit to 64-bit architecture. This would at least some level of testing and checking, and quite likely some amount of rewriting.
Even that mythical "perfect" code has a maintenance cost.
Re: Why does protocol software need "maintenance" (Score:1)
Well, no. If the code no longer compiles, someone needs to fix it.
It might be due to external factors, but it still needs maintaining and, often, requires some understanding of the codebase to do it safely and to test it properly.
So, whilst it may not need a full-time, active maintainer, it is not zero cost, either.
Re: Why does protocol software need "maintenance"? (Score:2)
The big difference between Windows & Linux is, with the exception of GDI & Twain, Windows can generally be coerced (sometimes, with a little prodding, registry-stuffing, .inf-editing, and/or self-signing the driver with a made-up key) into using even truly ancient binary drivers, as long as they were written for NT, Win2k, XP, or later.
In contrast, you can't even use binary drivers compiled on a different computer for the same fsck'ing version of Linux unless the build environments were absolutely 1
Re: (Score:2)
and a HUGE problem for devices like phones where Qualcomm just isn't ever going to voluntarily open-source its drivers
The lack of open source drivers is the problem. The linux kernel is GPL not LGPL, there is no reason that all those drivers shouldn't be open source.
That will push them to bsd/other kernel you say? Well so be it, maybe they'll actually adhere to the license conditions.
Linux is made for open source drivers for many reasons, but one of which is to avoid the very situation you're talking about where you can't compile/update older drivers for a newer kernel due to lack of source. Everyone gets the freedom to up
Re: (Score:2)
and a HUGE problem for devices like phones where Qualcomm just isn't ever going to voluntarily open-source its drivers
The lack of open source drivers is the problem. The linux kernel is GPL not LGPL, there is no reason that all those drivers shouldn't be open source.
That will push them to bsd/other kernel you say? Well so be it, maybe they'll actually adhere to the license conditions.
Linux is made for open source drivers for many reasons, but one of which is to avoid the very situation you're talking about where you can't compile/update older drivers for a newer kernel due to lack of source. Everyone gets the freedom to update/modify.
No, it's not. There's TONS of out-of-tree, opensource drivers for obscure hardware that are not maintained any more and won't even *compile* with modern kernel. So it's not only ABI, it's API as well. And no, "it's opensource so you can just fix it yourself" is no excuse for horrid design which does not bother to maintain backward API compatibility. Any userspace library which did that would quickly fall out of use because the developers would just move to something more stable, opensource or not. Why the h
Re: (Score:3)
So it's not only ABI, it's API as well.
Considering it's the kernels own internal API, yes, yes new kernel version will have changes to it's own internal API's.
Any userspace library which did that would quickly fall out of use because the developers would just move to something more stable, opensource or not.
The kernels userspace API is probably one of the most stable things ever. What you're asking is more like that client libraries update but don't dare change any of their own internal interfaces that are never externally exposed.
You want to write your own patches to a library, keep them out of tree and then complain when the git pull of the latest version won't cleanly merge because the dev
Re: Why does protocol software need "maintenance" (Score:2)
As I understand it, the main architectural difference between Windows & Linux is, the Linux kernel is kind of like a huge C program whose source is auto-generated from the "master" source by the build system, preprocessor, & macros... and the generated source is kind of like what you'd have if everything used hardcoded pointers to internal kernel data structures directly.
In contrast, the Windows kernel (or at least, the part that drivers touch most directly) is like a C++ or C# library where almost
Re: (Score:2)
Outside of graphics hardware, it seems like most hardware is just a dumb interface and the driver does all of the actual work. Bigger profit margins for the specialty chip makers to offload to the CPU for everything too.
Re: Why does protocol software need "maintenance" (Score:2)
there is no reason that all those drivers shouldn't be open source.
Sure there is... it's called "market failure". As a practical matter, if you want an Android phone that can do 5G on American networks, your chipset vendor choices are basically Qualcomm, Qualcomm, Qualcomm, Qualcomm, and... (... suspense-building drumroll ...) Qualcomm.
Qualcomm has no intention of open-sourcing its drivers, and American consumers have zero influence over Qualcomm's business policies, because they have total "take it, or leave it and do without entirely" monopoly power over US 5G.
That's the
Re: (Score:2)
Yup, the particular version of GLIBC has been annoying me on Linux, despite incompatible versions both being named libc.so.6. Ie, ignore the
6" and run "ldd --version" instead to get the real libc version. So I get the oldest system I can (centos-7) and build there. And still someone will be on a system where it won't work and wonder why I'm not actively trying to fix it for them.
(For a time people were statically linking some executables but that caused its own set of issues)
Annoying for me (Score:2)
I'm working on porting the Essex MUD-1 code to Linux, so need as many DECisms as I can have, even if I never actually use them in the end.
Re: (Score:2)
I'm working on porting the Essex MUD-1 code to Linux, so need as many DECisms as I can have, even if I never actually use them in the end.
That's cool. I spent a few late nights (fortunately only a few) in MUD-1 and 2 back in the day. Is the codebase online somewhere?
Re: (Score:2)
I asked Richard Barlte (one of the authors) for a copy and he obliged. It's truly ingenious, but a nightmare to read. I might well see if he's ok with me putting it somewhere where fans can work together to do a proper port. At the time I asked, my plan was to compile it in a DEC-10 emulator and see what happened, but - as he predicted - that's very hard going.
PatchWorks (Score:1)
The nickname we used for PathWorks was PatchWorks due to obvious reasons, but I give credit to the solution successfully scaling up to run Windows file sharing for many hundreds of PCs in the mid 90s.
In my office I run a couple of VAX emulators... (Score:2)
Re: (Score:2)
being anti-vax is too RISCy, these days. /// reused and recycled joke
Death to DECnet (Score:1)
VMS is dying (Score:2)
VMS is dying, Netcraft confirms it.