VM Enables 'Write-Once, Run Anywhere' Linux Apps

An anonymous reader writes "A startup will soon launch 'a kind of holy software grail,' according to an article at LinuxDevices. The dual-licensed technology is claimed to enable more or less normal Linux applications to run — without requiring recompilation — under Windows, Mac, or Linux, with a look and feel native to each. 'As with Java, Lina users will first install a VM specific to their platform, after which they can run binaries compiled not for their particular OS, but for the VM, which aims to hide OS-specific characteristics from the application. Lina comprises a platform-specific application that virtualizes the host PC's x86 processor... A lightly modified Linux kernel (2.6.19, for now) runs on top of the VM. Under the Linux kernel is a filesystem with standard Linux libraries modified to map resources such as library, filesystem, and system calls to analogous resources on the host platform.' Further details, including an entertaining video or two are at OpenLina.com"

Lofty Goals Indeed

FTA

"We're trying to change the economics of the software industry. Our hope is that in the next two years, all these companies that are starting with Microsoft APIs will start with Linux instead."

That is a key mindset to getting Linux out of its niche and into the mainstream. Thinking Linux first can't be a bad thing at all.

I hope this lives up to its hype (and promise). I may have to finally break down and get an Macintel (much to the chagrin of my PPC army).

This won't be useful for a MAJOR market segment.

Namely, games. I see nothing on their FAQ or Features pages about 3D support. Without that, games-- and several subclasses of applications (CAD programs, simulations, scientific visualisation programs, etc.) will fall flat on their face.

This is a noble effort, though, but what about the 3D?

Re:This won't be useful for a MAJOR market segment

Please :)
Are PC games crucial in every situation? I think it's naive to believe that there can't be a success for a technology just because it means it doesn't apply for demanding 3D games. I'm sure they can live without PC gamers and focus on the multi-billion dollar companies who want their applications to work seamless no matter the operating system.

Re:This won't be useful for a MAJOR market segment

True enough. But it's just like porn-sites and the defence industry and velcro, if you get my drift: 3D games have a tendency to produce offspring in 3D rotating multiple desktops and those quivering windows when you move them. And _that_, my friend, a user can never do without.

Re:This won't be useful for a MAJOR market segment

5% of a market of hundreds of billions of dollars is meaningful, whether you think so or not.

-jcr

Humor

From the LINA FAQ:

Q: What does LINA stand for?
A: LINA is not an acronym.

Re:Humor

I've heared they renamed MIT. It's now named "MIT Institute of Technology"

From the FAQ

Q: What programming languages is LINA written in?
A: For performance reasons, we've written LINA in C and C++.

Why not just write the VM in Java, then it'll be truly portable, right? Right..?

Business case hurdles

>"It's been said on numerous occasions that the big things that are holding Linux back are Adobe [for
>graphics software] and Intuit [for its personal finance software] because they write to Windows, and
>people can't give up those two pieces of software. As soon as companies like that, or companies that
>want to compete with them, start writing to LINA, things are going to change fast."

And how is the likelihood of big software houses starting to write to LINA bigger than big software houses staring writing to Mac or native Linux?

CoLinux

It's like http://www.colinux.org/ [colinux.org], but multi-platform. It really does not feel like revolutionary stuff...
CoLinux lacks only a good and painless frontend, and transparent support for X which would create new Windows instead of a X Server client containing Windows. The MacOS port does not feel like big stuff.
The only point of this VM would be performance.

Native Look and Feel

run (...) under Windows, Mac, or Linux, with a look and feel native to each

People should realize that this is impossible to do. This sounds like look&feel was something like a skin you can change at will, but it is not. Different systems have different requirements that cannot be changed programmatically, like keyboard shortcuts, icons, file placement, GUI metaphors and so on. If you want an application to feel native, simply design the GUI according to local customs, there is no other way.

Re:Native Look and Feel

Trick maybe the wrong word - changing the L&F is by design with code such as:

try
{
    UIManager.setLookAndFeel(UIManager.getSystemLookAn dFeelClassName());
}
catch (UnsupportedLookAndFeelException ex) {}

While not perfect, it is pretty good. There are screen shots and some of the deficiencies laid out here: http://today.java.net/pub/a/today/2003/12/08/swing .html [java.net]

Re:Native Look and Feel

People should realize that this is impossible to do. This sounds like look&feel was something like a skin you can change at will, but it is not. Different systems have different requirements that cannot be changed programmatically, like keyboard shortcuts, icons, file placement, GUI metaphors and so on. If you want an application to feel native, simply design the GUI according to local customs, there is no other way.

That's one thing (and you're totally right). And second:

As with Java, Lina users will first install a VM specific to their platform, after which they can run binaries compiled not for their particular OS, but for the VM, which aims to hide OS-specific characteristics from the application.

Which left me thinking... "and unlike Java.. it does what?".

We've got Java, which has matured over the years, we've got .NET / Mono. And we got Cygwin, which allows you to use the Linux API on Windows.

Looks like that startup has agenda to brings more of the Linux API to Windows, and thus help Linux become more mainstream as a development platform. But agenda does not business or money make. They've entered a crowded market, offering little new except impossible promises for native look and feel.

Why...?

"You have to compile binaries specifically for Lina, but it's fairly trivial, no different than compiling binaries for SuSE or Red Hat."

So it's like UML on Windows using Qt/Windows and Gtk directly? And it runs as a layer on top of Win32??

I'm not sure that this effort is really worth it if you've to recompile. With Qt4/KDE4 more or less all of KDE will operate on Windows. Most major open source applications are already functioning in Windows.

No matter what this thing does, it's still an extra layer between the win32 subsystem and the applications.

It's not truly transparent

->...you have to compile specifically for it, so it doesn't run legacy linux applications.
http://www.linuxdevices.com/news/NS6279947776.html [linuxdevices.com]
"In Lina's case, the VM is essentially a Linux environment that supports standard C/C++ applications, or even perl and python, if their respective interpreters are installed. CTO Nile Geisinger explained, "You have to compile binaries specifically for Lina, but it's fairly trivial, no different than compiling binaries for SuSE or Red Hat."
-> how is this better than cygwin/mingw???

Even worse:
"Open source developers will be able to use Lina for free, while commercial developers will pay an as-yet undecided licensing fee, the idea goes."
->so, better recompile for free for the three systems.

->wine is the other way round, but at least it doesn't need you to recompile or require you to pay to use it.

->no comments:
"Geisinger hinted that Lina's library set is fairly extensive, after four years of development by a team that has ranged from two to five developers. "There's a lot of code there," he said.
However, a few biggies are missing. GTK+ support is in the works, but not finished yet. There's no support initially for USB peripherals and possibly for other hardware interfaces. And, there's no slick installer to put non-Linux users at ease."
->compare with the resources put e.g. behind java or even cygwin

 

Enterprise what?

These kids should focus on making applications work on the desktop first. Hoping their C container will run multi-threaded multi-node applications in the enterprise is a whole 'nother ball of shit. But they don't know that yet. Java can barely keep up, and they've been at it for a decade. Contemplate delusions of grandeur after you get this to run something useful on the desktop. My advice. you'll just get your ass sued off otherwise.

Dual licensed - wtf?!

From their FAQ:

      'A: LINA is dual licensed. For non-commercial users, LINA is available under the GNU General Public License, Version 2.
              If you wish to use it commercially, please contact us to find out more about the LINA commercial license.'

Erm I'm sorry?! You can't stop someone using a GPL licensed program for commercial use.
Do they mean to say that if you want to sell it or do none-free changes then they will sell
you a non-GPL license?

Re:Dual licensed - wtf?!

There are several development tools that have such a license. That if you are developing GPL software then the GPL license applies, but if you are developing comercial closed source then you pay.

Though this is a Virtual Machine, not the software that runs on it, you typically have to have some sort of clearance or approval to install and run such software in a large company. And having a paper trail of purchase or licensing fits the traditional business model. So yeah, if they actually produce, it makes since, however...

How long will it be before someone gets the idea to simply put it all on a memory stick and/or live CD/DVD such that they can take it with them, yet run it on top of other running OS's, that would then allow the best of both worlds (let the best app win regardless of OS)....

BTW, a windows version of D-BUS is being developed and in a case like this, it can be what integrates the two system applications. And that is what will make it really useful.

Imagine taking your personal tool box to work that you keep and which adds to your value to the company.
   

Qt?

So it looks like right now it mostly supports Qt with some gtk stuff coming along. Anyone else find that odd? Today you can compile your Qt apps on Linux, Mac and Windows and get native look and feel. Why would I want to wrap that with a vmmachine? Just yesterday I ran across an app written in Qt for HDR imaging [sourceforge.net] that is written with Qt and is for the mac, linux and windows.

Don't forget wxPython!

You can also use wxPython (http://www.wxpython.org), which seems to start a lot faster than LINA (look how many times the LINA app bounces in the dock before it starts), actually *comes with* Mac OS X Tiger, uses native OS controls whenever possible and as of 2.8.3 has a library called SizedControls which automatically applies OS HIG-compliant sizing and borders to your windows and controls on Windows, GTK/GNOME, and Mac (disclaimer: I'm the author of said library). Plus, unlike Qt and LINA, wxPython/wxWidgets is free for commercial development as well as open source.

So I've been using this 'holy grail' for years, but maybe the VM slowdown and commercial licensing will appeal to some people. :-)

Just what we need...another VM..

Imagine a world where everyone wrote software for virtual machines - the problem occurs when people write software for different virtual machines. Eg: if Pidgin was written in Java, Firefox in Mono and GNOME in Python, to run my ordinary desktop I would suffer the overhead of 2VMs and an interpreter.

Besides, its not hard to write cross-platform C++ code.

Re:Just what we need...another VM..

VMs should not exist. Source-code level compatibility is more than enough to make cross-platform applications. The only benefit of a VM is run-time optimizations, as the code is translated.

VMs exist because no one dared to a make a C++ like language that guarantees source-code level compatibility in all platforms and has garbage collection.

And the premise "write once, run everywhere" is totally flawed. If you have N architectures, you need N virtual machines. If you go the direct route (i.e. no VM), then you need N compiler back-ends. Since the virtual machine is actually a compiler back-end (only executed lazily), I see no benefit from using VMs.

Dependencies?

I don't understand how it deals with dependencies, especially for GUI applications for that "native L&F." I could understand statically compiled binaries, but it obviously must use some shared objects on the OS because in the introduction video, Windows still required Cygwin.

I don't doubt that this will be useful, but there's just too much hype surrounding it right now, and I can't tell the difference between the truth and the embellishments.

'Write-Once, Debug Everywhere' Linux Apps

'a kind of holy software grail,' to enable more or less normal Linux applications -- without requiring recompilation -- under Windows, Mac, or Linux, with a look and feel native to each.
Just what i have been waiting for...

So what! Windows already has this.

Most apps under windows can run on 98, NT, 2000, 2003, XP and Vista!

"Look and feel native to each"

This time, for sure!

Sorry, that claim has been bullshit for decades.

-jcr

Deja Vu all over again!

A startup in Alameda, Calif. plans to release a kind of holy software grail

VM enables "write-once, run anywhere" Linux apps?

Ultimately, Lina has some pretty lofty goals.

Hi.
1. This sounds like what Java tried to do.

2. Startup!?!?! Watch your pocketbooks!

write once, test everywhere

If you compile different versions, or if a machine automatically translates for you, that doesn't mean you don't have to test on different platforms. If you expect to have a robust product that runs on linux, windows, and mac, you have to test it on all 3. I think people are confused that this will somehow eliminate that step, so you'd save yourself some time. If it's all one source base, then you'll have tons of stuff like this:
-if running mac, then do this fix, if running windows then do some other fix, if running linux then do some third fix
so either your code gets very large and unweildly, or you have 3 different versions and let them branch a bit. Either has advantages and drawbacks, but neither is what VM promises in theory.

Remember: "in theory, there's no difference between theory and practice, but in practice, there is."

Software Virtualization

While this is nice, personally Id rather see SVS or Softgrid get extensions to support non windows OS's. That way you still have your 'enterprise' management tools.

What business wants to adopt something that doesnt have management tools these days? I know i dont want to make more work for myself.

For the love of God, why?

I can understand the hack value, but why, for the love of God, would i want to run binary Linux apps on Windows? Didn't they have anything better to waste four years on?

There are some binary Windows apps, which could make life easier (albeit somewhat unethical in FSF terms) for Linux users, such as MS Office and IE6, and AFAIK that's what WINE is for (although i've never had the dire need to actually try it). But vice versa??

All the FOSS Linux apps that are source portable - OpenOffice, Perl, Mozilla, SVN, Audacity etc. - already found their success on the Windows platform. Is someone weird enough to make an application which is binary-only *and* Linux-only?

Or am i missing something?

fat bunny?

I'm not sure I want to use a piece of software with a fat bunny as its mascot.

MS should buy this immediately

If MS had half a brain they would buy this company immediately! This could work like wine in reverse! MS could say to the suits, "you don't have to migrate away from our wind-turd OS, you can buy this emulator!" The suits are too stupid to realize that they would be missing the most important advantage of Free Software, namely freedom, but the suits would like not having to convert their powerpoint presentations to open office!" This is horrible news!

Write once... debug everywhere!

It will end up with a smaller market share than Java with all the same problems. Debugging will be hell and the application on all platforms will suck because it does not co-exist well with the rest of the system. At least with java there were significant security benefits and a large enough base to make simple graphical tools useful. Extensions to java such as SWT also went quite a way in creating a good native UI widget set, but it still sucks.

Write Once, Run Nowhere seen in market again

I'm sorry, but the number of times I've seen the "Write Once, Run Anywhere" claim made for new "paradigms" is fairly scary. There are limitiations to all of these approaches. In this case, running software in complete OS emulation mode denies access to hardware features that have not been successfully ported to the virtual environment, enforces limits of the particular underlying VM hosting operating system in fascinating ways, and absolutely punishes the performance of any disk-accessing operations.

There are uses for virtualized environments, but they're hardly a new approach to code portability.

Can anyone say "pcode" ?

The claim in the article was trivially possible back in the days of CP/M, minis and mainframes: Pascal was compiled to p-code and interpreted on eveything from na 8-bit Z-80 to a 80-bit CDC.

I daresay we've improved on that since, with perl as one obvious example.

--dave

nothing new, but could be great if done well

If they make this thing slick and seamless, i'll be impressed. the technology is nothing special (something most OSS people sadly miss the point of ) it's the overall usefulness that's important.

UML, coLinux? line.sf.net?

Did they port UML to Windows? Are they using coLinux.

Why start at the kernel level? Why not start at the application level,
like line.sf.net did?

Yea but Java's soon GPL.

If Java's GPL then why do we need another Java?

wtf? Barring commercial use of GPL soft?

From http://www.openlina.com/faq.html [openlina.com]:

Q: What license is LINA released under?

A: LINA is dual licensed. For non-commercial users, LINA is available under the GNU General Public License, Version 2. If you wish to use it commercially, please contact us to find out more about the LINA commercial license.

How can they expect to bar commercial *use* of Lina when it's a GPL'd software product--unless their software must be embedded in the end executables?

Meanwhile, the video describing Lina is terrible. It shows (in a ridiculously puny window) two people installing an Apache-backed *WEB APPLICATION* onto two apparently different systems: a Linux machine and a Mac machine. What's the point of that when a PHP-backed application will do just as well and is nearly as simple use? (And what needed Lina? Apache? The web app itself? Both? Beats the hell out of me.)

There's no word on actual performance of Lina binary applications either, and while they claim additional "security," the reality is that complexity does NOT breed security, and Lina is yet another layer which must be maintained, secured, configured, and reconfigured.

Java already provides all or nearly all of this, and targetting development at Lina would be a massive re-tooling. It would also appear that the LINA PDF is internally inconsistent on the matter of whether legacy *binaries* or just legacy *apps* would run under a Lina host. I'll guess that everything must be recompiled specifically for Lina for it to work properly.

Quite frankly, once Java's GPL'd code is ported to the missing OSes it needs to be ported to, there will be no barrier to Java adoption anymore. Plus, commercial devs can still create independently-licensed Java applications without worrying about Lina demanding their cut for commercial development under Lina.

VMware without the interface ?!

If I read the article right, this is basically VMware (or Bochs) with a different name. How is that going to change the world again ? Don't most distributions already offer binary packages in the first place ? Just because I run Gentoo doesn't mean my mother has to.

For all I care, it'd be nice if we could cull the hundreds of vanity distros and repatriate all those developers onto Ubuntu, Red Hat, Gentoo, Debian and maybe Slackware (if they even care). Heck we could probably merge all the binary distros into a single one with "profiles", because typically a binary that runs on one standard distro, will run just as well on another distro running on the same architecture (as long as its dependencies are met). In a pinch, I can build something on my Gentoo, copy it over to my pain-in-the-ass CentOS box and run it. x86 is x86 no matter what crap you load onto it, it still runs x86 code.

More? Or Less?

"... its dual-licensed Lina virtual Linux machine will run more or less normal Linux applications ..."

Well, what is it? Is it more, or is it less?

Requesting a Dupe

To: Slashdot Editors
Re: Dupes

I know we bitch about dupes, but can we get a dupe of this article when they actually RELEASE this please?

Thanks!

Bunny!

I will use this product and/or service because I like the bunny.

That is all.

I won't comment on vapor-ware

Am I "net-blind" or is there no where to download this thing? Google around for "Download Lina" produces hits on some MP3s and some hits on the Lina logo. If there is a download version of this project I can't find it.

On the matter of the product itself:

From the videos on the website it sounds like Lina is a new type of VM for C/C++ code similar to the JVM. If that is the case I can easily contain my excitement and hope these guys have a good marketing department. The technology is not 10x better and as such will have a long road to success.

If you can write to Lina and distribute Lina "binaries" or native binaries (that is "exe" on windows) I don't think you'll see much resistance to Lina as a product. You may not see much fervor over it either though.

If you can recompile a Linux project and distribute either "Lina binaries" or native binaries for a program then I think we have gold here. I'll be very excited and it means that Linux could morph into a kind of super Java style API for all Operating Systems... a sort of meta System V.

If all Lina does is provide a VM to write to, (which is what I suspect), then Lina's success is going to be a matter of marketing. But not traditional big marketing... it will succeed on a combination of smart technologist marketing and viral marketing. If Lina has that it might carve out a niche for itself.

The question is, true believer, can Lina make your heart flutter like Linux, Ruby, Python, or PHP did? If it doesn't learn to make your heart go pitter-patter then I it will have to find a way to cut deals to make you want to learn to dance with Lina. Maybe folks at dLoo can cut a few deals that move a critical mass of developers over to the Lina side of the force.

I don't know if I like Lina or not, I haven't even met her. With a hook like this on Slashdot today would have been the perfect time for me to meet Lina and maybe have dinner. As it stands, I'm a desirable geek and get lots of young new technologies interested in my time and attention. I've got Beryl that I'm hanging out with right now, some python code from Numenta that keeps calling me back and looks mighty nifty, I've got new FX-y tech I'm going to spend some time with too... And, that's just this weekend. I may not notice Lina again. I'm sure she's a nice girl with great personality but... the other tech I can meet and talk to right now and Lina didn't even give me a month and day to get back to her on.

Why not Java?

I don't know which "Linux APIs" they standardize on in their product, but there are very few "Linux APIs" I would be interested in standardizing on as an application developer.

Perhaps this has value for companies wishing to port legacy GNU/Linux products to Windows and OS X with less effort than is required to target Cygwin + OS X native, but then how many companies are there in that situation who are willing to accept the risk of targetting a brand new VM platform from an unknown vendor, and do not already have the inhouse expertise to maintain Cygwin and OS X ports?

And why not Java? That's the question they need to sell any intelligent prospective customers on. If you care about performance to the extent that you don't want to target Java (VM) in the first place, why would you want to target this VM virtualizing the entire Linux kernel, a VM that doesn't have the benefit of years of tuning and performance enhancing technologies from some of the brightest engineers in the industry (Java)? And once again, a VM from an unknown player that is not proven in production?

My suspicion is that this is just a bunch of people looking to be bought out, or at least hired, which is a perfectly reasonable thing to do, ...unlike, perhaps, basing a new product on their technology.

This sounds like a reverse WINE

WINE runs Windows software on Linux.

LINA runs Linux software on Windows.

I guess originally they would have called it LINE, but then changed the last letter to A to avoid being just like WINE?

Part of the reason why people develop for Windows are that the development tools are easier to work with than Linux and other operating systems, like Visual Studio, Delphi, etc. Now you have to use Linux development tools, I am guessing C and C++, and Linux libraries to make code to run on this LINA virtual machine based on Linux APIs and the Linux Kernel? The best thing you have in Linux next to Visual Studio is the Mono project, but the IDE needs a bit more work, and the Winforms translation library needs more work as well.

I would like to try it out sometime and see what languages it supports for developing under LINA, when one does not have a Linux box to write the software on. I wonder if it supports the Python language and allows installing other languages on it, so we aren't limited to just C and C++ development.

Stupid

On windows this program requires Cygwin. So yes you can run all these apps natively, as long as you first install a extraction layer. And hey I wonder if I can get Cygwin working under wine so then I can go through two extraction layers. Wait, maybe I can then install colinux in wine that pumps x output through cygwin, then I can install wine, and then get cygwin running, then I can install cygwin...

Just write a fucking app in good c/c++ and staticly link libraries not on windows. Compile it and wohoo, a binary for windows. The only thing the least bit interesting is the gtk/qt to native api layer. That should be the library they provide developers with for ease of compiling to different OS's.

Java + SWT

How is any of this doing anything at all that Java using the SWT api's does not already do?

cygwin?

Obviously these are 2 different approaches, but anyone care to give som pros and cons of the cygwin approach and this new approach? they both require recompilation so im wondering what the benefit is?

is there no "cygmac" or equivalent?

WORA is a Myth

Write Once, Run Anywhere will never happen for two reasons.

First, there are different utilities and features available to each operating system. Every developer that ignores these features and utilities in order to write a program once for all systems will be at a disadvantage relative to the person that writes a native app. In other words, the native app will run better than the generic app.

Second, people gain access to a flexible choice of operating systems, but the developers get locked into a proprietary language/framework. Again, limited choices means that for some (many?) problem domains, a different language/framework will be a better choice, and by denying themselves that choice, the developer loses a competetive advantage to another developer.

And that's ignoring the issue of just getting it to work as advertised... Java hasn't even gotten there, 15 years later. I'm sure a small subset of problems will work just great with Lina... but I don't think it will do any better than Java in the long run.

Raven

Objectification

When will this ever end?

1. We had assembler. It wasn't portable we had to recompile and rewrite for different platforms. Libraries
   , (to make things easier and for 'sharing' code), were obscure pointers to areas of memory.
   
2. We had C. It was better at porting, but we still had to recompile for each platform. No obscure assembler
   that required a rocket scientist to figure out what the hell was going on. Wow! We had 'functions' - we could
   share code a lot easier!
   
3. We had C++. Much better at porting, but we still had to recompile for each platform. No obscure assembler
   that required a rocket scientist to figure out what the hell was going on. Wow! We had 'objects' that virtualized
   concepts. But why do we suddenly have 'fat' programs.
   
4. We had Java. No issue at all with porting. No obscure assembler that required a rocket scientist to figure
   out what the hell was going on. We had objects that virtualized everything. 'fat' programs? - forget fat these
   were obese.
   
5. We had virtualization. No issue at all with porting. No obscure assembler that required a rocket scientist to
   figure out what the hell was going on. Forget trying to virtualize program space - heck let's virtualize the whole
   damn O/S! What's a fat program without a fat O/S?!
   

So, what's next after O/S virtualization? We've tried in the past to objectify and virtualize program space
and to a large extent doesn't work as we either keep changing userland requirements or our methodologies force
us to change.

Don't get me wrong - love virtualization for all the right reasons. But, all we have at the end of the day
is faster and faster machines that sit there chewing up greenhouse gases, (Bring back the old days of the
TRS80!).

Don't we already have web apps?

This really sounds like they're trying to fill a problem that doesn't exist. If you need apps that satisfy the lowest common denominator, then you build a web app and suck up the time for web development. If you need a highly performant and responsive app, then you write a platform-specific forms app.

Running the whole thing through a VM just wrecks performance and still requires massive testing, plus you lose all of the OS-specific features b/c you had to code for an OS-agnostic VM. Even if it works, it's still wrong :(.

another cygwin

from their article this sounds like another cygwin... only they've added virtualization (so like cygwin but slower).

The main feature seems to be that GTK and QT will be mapped at runtime to win32 APIS... which is... ah... an interesting choice since GTK and QT already run on windows.

This doesn't sound like something that will have mass market appeal... although some companies that want an easy way to port existing unix tools to windows (if they are converting to windows) may take advantage of this.

I have no conception of how the authors imagine this will spur linux adoption. If this works right, the user wouldn't be aware they were running a linux environment. If it doesn't work right (like cygwin) users will be annoyed that they are using some big clunky vm that doesn't integrate with their os.

P-Code ... again

Are we really stuck in the loop of thinking that interpreters are as good as the real thing? We had BASIC in the 1970's then we had Pascal with its P-Code in the 1980's then we had exactly the same thing only it's called Java and a JVM in the 1990's. We are stuck in a rut - none of this is new just the structure (and name) of the language changed.

rPath

This seems like a more mass market application of rPath's idea of building custom VMWare images for specific applications (although the VM sounds like it will be less distinct from the host OS, hopefully this will give performance increase). It is a pretty clever idea, and essentially flips these kind of ideas https://wiki.ubuntu.com/SeamlessWindowsIntegration [ubuntu.com] on their head.

Linuxolator!

Looks like the Linux Binary Compatibility layner [freebsd.org] of FreeBSD [freebsd.org].

Re:Huh?

Why are people afraid of recompiling? It is pretty painless if the source is packaged well.

If you can get my mom to understand that sentence, I will pay you $500.

Also in how far is this groundbreaking? Seems to be a pretty forward and not too complicated engineering task.

Interestingly enough, the word 'groundbreaking' was not used even once in the summary or the article. News doesn't have to be groundbreaking. It could be a very simple old idea used in an ingenious way to be a very useful tool for the masses. Like this, they aren't hiding that they're kind of copying what Java does. But, you know, if it was such an easy engineering task, why haven't you done it? I'm very interested in where this goes.

Re:Huh?

The users don't have to compile silly.

With something like Qt, its "write once, compile three times". Distribute binaries for Linux, OS X, Windows. All with a native look and feel, using native compilers and libraries.

Its not clear how this thing works with the GUI. Is it a new toolkit? Is it a hack of the Gtk toolkit? (I thought it was funny how the demo talks about the "native Gtk looknfeel from within what looks like KDE).

Re:Huh?

"Dear user: Insert the CD. Type make all; make install. Press return and go for coffee."

It should be

"Dear user: Insert the CD. Click 'Install'. Click 'OK' and go for coffee."

See the difference from a user's point of view?

Re:Huh?

Insert CD.
Phone relative because the CD does nothing.
Find Install and click it.
Cancel the dialog and click the other install.
Phone relative again and ask them why its going to take 3 hours.
Make coffee.
Return to computer and switch it off (thinking they were switching it on because the screen was blank)
Ring relative and ask why its not worked.

Re:Huh?

Actually, most Windows apps take advantage of autoplay, which means it really is "put the CD in, push the big flashing button that pops up, click next a whole lot because none of that text could be important, wait".

To my knowledge, Linux doesn't have autoplay. While I agree that autoplay is awful, it does make things easier for endusers.

I don't know if OSX has autoplay or not, but in any case with OSX it tends to be "put disc in, double-click the disc icon that just showed up". I haven't seen any equivalent for Linux - you usually have to find the install program or similar. God help you if you have a package manager - then you have to search for what you want to install!

People's brains freeze up when confronted with a computer. They'd much rather just put a shiny disc in and let the magic computer do its work. Seriously, you could sell "Linux application install discs" which are just a pack of CDs where each one has "aptitude install gimp", maybe a hudnred bytes each, and people would buy them.

Re:Huh?

Considering I've never ever installed a program from a cd under Linux dont you think autoplay would be useless? :P

I think package management is far superior to cds. For one you cant lose your package management and spend half a day looking for it.

Re:Huh?

Most people, I suspect, want plastic discs to install software off. It's far easier conceptually to deal with. Linux really doesn't handle that case well.

Personally, I agree - I vastly prefer package management or online downloading. I suspect there will be some user re-training needed to get most people to understand that, however.

Re:Huh?

Ah, you've used Windows, I see. Installing things is much easier on pretty much every other platform (including macosx, most of the linux-based platforms, and most of the BSDs).

What are you talking about? One of the big problems in the Windows world right now is that applications are so easy to install, that the user often has to run special software just to keep applications them from installing all by themselves!

Re:Huh?

install.sh
----------
#! /bin/bash
make && make install

What was your point, again ? Oh yes, there is no "Click 'OK' " step, do you care ?

Re:Huh?

I'll say it very simply.

Do not force end-users to the command line.

Re:Huh?

I'll say it very simply.

Do not force end-users to the command line.

This mentality bothers me. While the command line is intimidating at first, and end-users should never have to learn how to navigate the command line, if I'm giving instructions, I would much rather have someone using a command line.

For example, a few months back, my girlfriend wanted me to put Linux on her computer. She was saving up for a Mac, and her anti-virus had expired on Windows. She needed a web browser, office suite (She used OpenOffice on windows to begin with), and an instant messenger. I had her install Kubuntu, answering a few questions when she had them. Once it was installed, I pulled up a terminal to start installing some programs and codecs with apt. She was deathly afraid of learning the terminal, so I started stepping her through the installation with Adept Installer. The instructions for installing Flash went something like this:

- Click the "K"
- Click add/remove programs
- Type your password
- Check the box next to "unsupported"
- Check the box next to "Proprietary software"
- See where it says "KDE"? Click the down arrow and select "Any Suite"
- Type "flash" in the search box
- It's not in multimedia? Try "Others" I guess.
- Check the box next to "Macromedia Flash Plugin"
- Click "Apply Changes"
- When it's done, click "quit"

Alternatively, I could have told her:
- Click the 'K'
- Hover your mouse over 'System'
- Click 'Terminal Program (Konsole)'
- Can you remember that? Next time I may just say "Open a Terminal"
- Type "sudo aptitude install flashplugin-nonfree"
- Type your password

I explained to her that I didn't expect her to learn how to use the command line on her own, but it's a lot easier for me to tell her a command when I'm giving instructions. She hasn't used the terminal once on her own, and she's enjoyed perusing the programs available through Adept Installer, but she knows if I have to give her an instruction, it will be a lot easier to use the command line.

I realize it's initially intimidating for users to have to open a terminal, and I'd like to see graphical interfaces for everything a normal user would need to do, but I also wish we could get the average user to where they realized the set of instructions is a lot shorter when someone gives you a command than when they have to explain dozens of clicks.

Re:Huh?

Telling people how to fix problems fosters dependency on you. Showing them how to do it, in a manner that doesn't take years of command-line dorkdom to understand, is probably far more helpful.

Though in this particular case, if your girlfriend's about to get a Mac anyway, maybe it doesn't matter so much. I'd be more worried about getting dumped once she realizes she doesn't need you to fix her computer anymore. :)

Re:why Windows has a 'run' command

Actually, knowing how to spell hinders typing Linux commands...

Re:Huh?

This is slashdot. You lost me at "girlfriend".

WTF is that?

Re:Huh?

I would much rather have someone using a command line

And therein lies the problem. You would rather have them using a command line. They don't want to. When you have a GUI, you always have prompts and a safety net. You can say, "Click on the button that says 'Change Setting.'" They have an automatic double check because there is something that matches, exactly, what they're being told to do. They have limited choices because they're working from a menu or from components and when one matches what they're told to do, they have some kind of confirmation they're pressing the right button. When you say, "Now type this," then you can hit any number of snags and they know it. You can tell them what to do and they can use a single quote instead of a double one or hear the word wrong or mistype it and, from their point of view, they don't know what's going to happen if they make a mistake.

You are much wiser than most developers I've seen post here who want to blame the uses for not knowing everything they know and you've got a good point. You and I can often move much more quickly with a console, but for users, the mere thought of having to type commands is frightening. They're looking at a blank screen with no feedback until they hit and then it could be too late.

You've realized, though, that it isn't about what you or I want, but what the user can handle or take care of on their own. That's the problem: It's our job to give them what they can handle. That is rarely a console. With my clients, before I got my software up to the point I wanted it, I had them install RealVNC on their computer and used a tunneling program I wrote in Java so it would go through their firewalls, then had them add me, and I configured RealVNC so it would only run when they wanted to run it and so it had a strong password. (And before anyone starts screaming, I'm simplifying and leaving out discussions with their bosses and IT departments about safety.) When they had problems, I just had them run RealVNC and add me, then I could fix, in less than 5 minutes, what could take 45 minutes or more if I were telling them what to do. When I finally got my own program to where I wanted it to be, that wasn't an issue anymore since I had enough failsafes they didn't need that kind of help anymore.

Re:Huh?

Lemme tell you a small story here.

When MS released Windows 3 was when it started to become known in Brasil. At the time, there was a TV show called "Confissões de Adolecente" (no need to translate, since the name is not relevant). At the time, there was an episode where the main character was bitching an complaining about having to use a mouse and click on stuff. After all, if he wanted something, all he had to do before was to type the command. Now, he had to search for the icon, click on it, than click on something else etc.

So, to translate your comment into something that really means something (and is actually true), what you mean is:

"Do not force end-users to do something different"

Here is Brazil we have a saying for cases like this, which roughly translates as: "If you change the color of the grass, the mule will starve to death"

Oh How I Wish It Were That Easy

"Dear user: Insert the CD. Type make all; make install. Press return and go for coffee."

Your attempt at simplifying building from scratch amuses me. First off, you put the period in your instruction type font which I'm sure my tech inept mother would type into the console. Which would result in:

make: *** No rule to make target `install.'. Stop.

And how would she know to open a terminal? Would she know how to open a terminal?

Second, what about dependencies, how does she know to read the README file or anything else to figure out what she needs to build this source. You don't exactly include all the source of all the libraries you coded with, do you? Rarely have I seen a project coded from scratch with no dependencies.

See, the issue here isn't that she can't be instructed like a monkey to hit a button. The problem here is that if something goes wrong, she's out of luck. I mean, as it is, the concept of double clicking what you downloaded to install it was a tough one to drive home. And even now I worry about her willfully installing viruses or malware on the home computer. Because she just doesn't understand the concept so well. When you ask someone to build from the source, you're pushing them quickly into something they don't understand and it's just going to result in a bad experience. The ease of use for software is actually more important to most people than its efficiency or anything else. Why do you think Java is so popular?

Re:Oh How I Wish It Were That Easy

Ok, what I gotta point out for you and whoever first brought up their mom in the whole compiling issue in the first place...

If your mom is technically savvy enough to have created an application that is now to be compiled from source to different platforms then she obviously understands the original sentence.. and the typo... Unless she is a programming savant who creates source code not knowing what it is.

Again.. the original poster was talking about a (developer of an app) (compiling binaries from source) (for different platforms), not an end user or your mom.

Re:Huh?

Given, I'm running a PPC machine, but from my experience with open source on Linux and Mac OS X... because something always, ALWAYS, goes wrong the first time you do "configure" or "make?" Always.

Either it's missing some libraries (my experience with GD), or it requires SUDO permissions but the instructions didn't say it required SUDO permissions, or the path its writing to is wrong, or it has a Good Ol American compilation error... something always goes wrong.

If you want to write a GUI to cope with every single possible error in the 'configure' and 'make' process, more power to you. But I doubt it's possible for any computer program to handle every case in an automatic fashion.

Re:Huh?

"Dear user: Insert the CD. Type make all; make install. Press return and go for coffee."

And come back to:

make: *** No rule to make target `all'. Stop.

And that's assuming the user knew enough to open a terminal and navigated to the appropriate directory, which you left out. Let's say after they got that error they look at INSTALL and discover your instructions forgot to include 'configure', and now they get:

configure: error: Enchant library not found or too old. Use --disable-spell to build without spell plugin.

Let's say they go ahead and disable spell-checking, even though it would be a useful feature. They type 'make install' again, like the instructions said, and get:

cp: cannot create regular file `/usr/local/bin/foo': Permission denied

Maybe they're not quite frustrated enough to give up just yet, and they do a google search and discover that you forgot to tell them to run as root. Hurrah! It installed! They type the application name and get

cannot open display 0:0

Whoops, still root. Maybe they realize this (smart user!), exit, and try running again as themselves. Oh, damn, the application installed in /usr/local and there's an old copy in /usr, so they end up launching the old copy instead. Even if that wasn't the case, most users are going to want menu entries and icons without having to set them up manually.

Re:Huh?

This is exactly why linux users get a bad wrap... wake up, seriously, if everyone was a tech guru you wouldn't have a job...

Re:Huh?

"download: install: run".

That's what we do! It isn't 1979 anymore and having to compile source code isn't something the average user should ever be expected to do....

Re:Huh?

Well, recompiling an application doesn't port it to a different OS.

The work needed to abstract the fact that you are running on Windows or MacOS instead of Linux is highly repetitive, and therefore a good target for factoring out into some common -- thing.

The most accepted way for this is to develop a framework with WxWidgets. But what if you don't like the framework? What if you need a different framework? What if your language is not supported by the framework? Integrating a VM to the underlying OS is an alternative.

Another thing that I think is useful in this approach is potentially dealing with coupling of unrelated applications via common library dependencies. If one application requires a later version of a library than another is compatible with, you can't run them both easily on the same machine. Anybody who used non-Ubuntu repositories on Ubuntu has run into this.

Re:gui needs a framework

Qt if you want a good one.

I'm sorry, have you ever USED a Qt program on anything other than Linux? I've yet to see a single Qt application that doesn't look and behave like ass on Mac OS. (That said, I've never seen a wxWindows one that didn't look and behave like ass on Mac OS either.) My personal favorite is RealBasic, since I've seen RB apps that look and feel native on Mac and Windows, but you'll never get the open source community to use it because it's proprietary.

Re:Huh?

Can you guarantee me that the source code will compile and operate on what ever the target system is? The point of the VM is that it virtualizes the system that it is running on. It's up to the VM on the host OS to do that guarantee. Once that is done, then all that needs to be done is target the application build to run on the VM, not one of dozens of different OSs or hardware platforms. Sure "Hello World" or your basic application may run fine when you just "make; make install" it, but for something much more complex, relaying on specific hardware or OS features, it may not work as intended.

Re:Is this going to hurt Linux?

Depends, if it takes off and developers write code for lina then by default those apps will be available for linux. Of course its a pipe-dream but if it became the defacto way to develop new applications its entirely possible that in the future those apps such as photoshop, etc would be able to run.

However it is already possible to write one application and deploy accross OSX, Linux and Windows. Simply write and compile against WineLib .

Re:Huh?

When you get as far as trying it, you'll find out that making the source compile on another operating system is 80% of the work in porting.

Re:Linux (non-)portability

I can't even get applications to run under different versions of Linux without recompilation, given the differences betweeen glibc and gcc over recent years.

I found that to be especially true in the 1998-2002 timeframe, but much less so today. Nowadays all the applications I've run across post-2002 will at least run on all the major desktop Linux variants with minimal hassle. However, embedded systems still tend to require recompiling to the target libc as they don't often ship with multiple versions of it.

Re:Linux (non-)portability

Please don't spread FUD, of if you're serious, please give useful details. Google and other companies have downloadable binaries, so I know it's possible. You should be able to develop on a machine that adheres to the Linux Standard Base and have that run other places without a problem.

A common newcomer mistake

DLL hell is a Microsoft only problem. There is no reason why you can't have more than one libc apart from a lack of disk space - it works by version numbers to avoid loading the wrong library. Your ten year old binaries will still run in most cases as long as you have the other libraries somewhere on your library path. Some distros have the old libraries neatly packaged as "lib*-compat" so you don't have to find them yourself to get an old binary to run.

Re:Huh?

You assume that I have the source code to recompile....

I run plenty of binary packages at work and it would be wonderful if the "just ran"(tm).

Re:Huh?

"if the source is packaged well."

see you just fucked your whole argument right there.

Re:Huh?

I've installed MythTV Frontend in a VMWare virtual machine that I run on my Windows XP laptop. It connects to my Linux mythbackend and runs great. I can run full screen and you would never know it was running in a vm. This capability may not make much sense moving from one Linux distro to another and it certainly isn't groundbreaking, but when you want to cross OS's, its great.

Re:Linux (non-)portability

Acucorp has been doing something like this since about 1985 with their COBOL compiler. It produces object code that is executable by a platform-specific run-time system; it is availabale for just about any platform (except IBM mainframes). The object code is copied to the target platform using a special utility, and it works. They do some spactacular stuff with databases, too.

Re:Huh?

Doesn't mono do this already? What do we need 50 VM's out there. I think this situation is getting ridiculous.

Huh? Storage and handling.

"Why are people afraid of recompiling? It is pretty painless if the source is packaged well."

Like in Squeak for example.

Re:Huh?

You've obviously never compiled all of KDE from source before.

Re:REALbasic?

So what popular software is written in it?