No more updates for Freecode [LWN.net]

No more updates for Freecode

Posted Jun 29, 2014 1:01 UTC (Sun) by khim (subscriber, #9252)
In reply to: No more updates for Freecode by roblucid
Parent article: No more updates for Freecode

We had LSB and almost NOONE cared, and I can't remember an LSB application.

Why would anyone care about that POS? LSB was never an answer, it was obvious years ago.

ABI's for purveyors of binary blobs, just aren't popular, distro's tend to support FOSS not closed s/w.

Yet even for FOSS it does not work. I could not run GIMP 2.8 in two years old Ubuntu 12.04 (Precise Pangolin) yet I could on decade old Windows XP. That's just pathetic.

It has to be possible for 3rd party to do binary right though given sufficient motivation, direct Firefox downloads worked well for me even if it's fatter than a well packaged native version would be.

Sure, but this is an attept to put shoe on the wrong foot. If OS maker wants to have a plethora on applications to choose from then it must make it easy to deliver them. Why would anyone try to target desktop Linux with it's insane fragmentation and minuscule market share if they could target Windows, MacOS, iOS or Android instead?

For a time the void was filled by people who grew up with large UNIX boxes. Linux was treated as "mini UNIX" and all was well. But over time people cared about it less and less and today it's suddenly finds itself with acute shortage of developers.

The commercial software mindset though, tends to target a particular system like RHEL/SLES and certify a version.

IOW: they are doing what they do with all other OSes. When program is sold it's sold for a specific version of Windows (or couple of them) and specific version of MacOS (sometimes few of them are listed). After that—it's responsibility of an OS maker to support compatibility. When new version of OS is released and new version of program is released thay may test with a few different versions of OS (today usually Windows Vista to Windows 8.1 is tested) but they still supply users with one set of binaries. It's only natural because it simplifies support. Why do you think they will attempt to do something else? But Linux distributions start foaming at their mouth when presented with such a approach.

Schemes to do an install time configure, via scripts & a dynamic module scheme must be technically possible, I haven't heard of one in operation.

You've not used nVidia hardware, then.

If FSF had been as inflexible about where their source built and ran, then Autotools would likely be insisting on SunOS4 or BSDi under some virtual Wine like emulation layer.

Right. But it's not an excuse to request such support from all other developers. FSF wanted to spread it's creatings as wide as possible. That was wize choice, sure. It means that these things are used everywhere and by everyone (after GPLv3 fiasco they are slowly replaced with BSD-based offerings, but scheme worked well before that). But if your OS only supports such packages and nothing else then it's DOA on a desktop. That's because desktop users actually want to use “latest and greatest” software when they hear about it—and not only superportable software like Autotools-supported GNU software.

Well, not all desktop users want to use “latest and greatest” software, some are happy with the destribution's offers. But when only these are left and everyone else switched to Windows and/or MacOS then FreshMeat/FreeCode becomes entirely pointless—that's where we've started the discussion, remember?

to post comments

No more updates for Freecode

Posted Jun 29, 2014 15:11 UTC (Sun) by roblucid (guest, #48964) [Link] (1 responses)

Philosophically, I think Linux desktop gets further marginalised (compared to turn of century) because many of the more pragmatic types, have been seduced to vendor controlled systems like OS X. Windows got it's act together on reliability over years, Win 7 is pretty solid, unlike NT4/2000/XP before Vista's WHQL driver model. So I think I agree with your last point, though right now I do indeed struggle to remember where this started :)

Frankly I'm having trouble understanding your argument, where you're going. It seems to be inconsistent and making some statements refutable by solid evidence, but perhaps it's an entertaining thinking out loud style discussion. So I'll try and be clear ...

Firstly you complained about distros and a lack of an ABI, so I mentioned the LSB.

Secondly, I left Nvidia out of it, as they use a kernel module shim for a driver for their hardware, rather than being application ISV. I have actually used Nvidia a long time ago and have avoided their GPU since, due to their refusal to document & support FOSS developers with information, so you're basically right about not using their GPU (though the old box I'm writing this on has an Nforce chipset).

LSB IS an ABI. ABI's are about boring stability and unchanging API's in sytem library, most change being bug fixes with occasional careful introduction of new compatible features aka Linux kernel. Some distro's supported it well and I've used the hooks provided to increase generality of my scripts.

But for a lot of Linux people "ABI considered harmful!", they want to develop and not have discipline of maintaining compatability, through carefully thought out and planned longterm change.

For example in yr 2000 with LSB, a week or so before LSB 1 was announced, RH 7 came out with an experimental gcc, which made C++ programs incompatible with both gcc-2 & the next release gcc-3. Market leader, basically desiring fragmentatory innovation over binary compatability, and FOSS was unaffected as C++ packages just needed an RPM rebuild. The 2nd-5th tier distro's had more interest in success of standardisation, but as the bubble burst pressure on balance sheets overcame their big market opportunity chasing optimism.

Now your beefs with LSB in previous thread about icons and stuff, I think you expected too much. Stuff like that gets handled by simple scripts. Those can be developed as FOSS with community support. Also in past, I have run very old binaries and they HAVE worked. But I wouldn't expect something like the GIMP to work, with the changing desktop environments and toolkits; unless your old GIMP had been built against bundled toolkit libraries. People just don't do that, when a distro is built, they want a lean mean package optimised and targetting their current release.

Having been in position of doing ports, build and distribution for an ISV, I certainly DID not need 100% compatability. When you have high value software, for decent money, then you can invest in glue at install time.

What you CANNOT do, is cope with churn and support difficulties, caused by all those little tinkering "improvements". Just look how difficult it is for MS to move userbase forward from XP with Win8, despite being in a position to dictate to OEMs and end users.

Desktop fragmenation is an issue for mass market S/W. From example of Apple and Google Android, that now it is about platforms, distro's are niche, hence Google Chromebooks, FireFoxOS, Steam etc. Hence punts like Unity which hope to break into new markets during a disruptive phase and stop Canonical losing money.

Right now, the Linux desktop is a play ground.. it's for developers, enthusiasts and those with deep pockets to speculate with, on the next latest & greatest "Big Thing".

Also, even if a distro goes out of it's way to support alien software, the OS/2 story tells us where that leads, ISVs not targetting native.

Distro's also with optional repo's and suppport for rolling release, I just feel you're perhaps ignoring efforts some make, based on perhaps a small sample size?

The real thing is, that lack of market means, there's no motivation to do what Firefox does and have knowledgeable ppl package for you. For FOSS authors, distro package maintainers take care of release details. So the curse of the "unusual" applies, more and more opinion is stated as fact and a technical solution is not adequate because it won't be accepted.

In the end, it doesn't matter if something is true or not, whether it's contra-factual, because many unequestioning people believe it, it becomes de-fact as real a block, as if it were true.

No more updates for Freecode

Posted Jun 29, 2014 17:50 UTC (Sun) by khim (subscriber, #9252) [Link]

So I think I agree with your last point, though right now I do indeed struggle to remember where this started :)

We started with the assertion that in a world of Linux distributions sites like ~~FreshMeat~~FreeCode are useless because distribution makers erected roadblocks between developers and users.

Everything after that were justifications for these roadblocks and attempts to explain how they are not a problem and how developers and users and stupid to not “see the light”.

Firstly you complained about distros and a lack of an ABI, so I mentioned the LSB.

Indeed, the biggest problems with the Linux distributions is the simple fact that they offer no ABI and no usable SDK to an application developers. LSB is just a fig leaf which is useless and pointless attempt to hide that fact as I explained may years ago.

But for a lot of Linux people "ABI considered harmful!", they want to develop and not have discipline of maintaining compatability, through carefully thought out and planned longterm change.

Indeed. Sometimes it looks like both third-party developers and actual users are some kind of obstacle on the road to nirvana.

For example in yr 2000 with LSB, a week or so before LSB 1 was announced, RH 7 came out with an experimental gcc, which made C++ programs incompatible with both gcc-2 & the next release gcc-3.

Actually this is what I consider proper and correct way of handling things. They needed a compiler with decent C++ support. Said compiler was not available back then (RedHat 7 was released September 25 2000, GCC 3.0 was released June 18, 2001, that's year and half after release of RedHat 7.0). Sadly they have not pursued the needs of desktop users with such a rigor.

The 2nd-5th tier distro's had more interest in success of standardisation, but as the bubble burst pressure on balance sheets overcame their big market opportunity chasing optimism.

Instead of offering usable, comprehensive set of ABIs to give to application developers they tried to pick bits and pieces which were already stable and use that as basis. It'll never work: something will always be in flux. In that case someone must decide that “we are supporting X using Y and W using Z and are ready to support these packages for the next KK years even if original developers will abandon them”. Ubuntu does it right in today's world, but, sadly, it does not look like they have enough manpower to create a coherent result.

Now your beefs with LSB in previous thread about icons and stuff, I think you expected too much.

Why? I don't ask for a “Visual IDE” or any wizards. I'm asking for the bare necessities: an SDK which makes it possible to create a package which I could give to my users. Nothing more, nothing less. Why else will anyone want to read my announcements (on ~~FreshMeat~~FreeCode or any other site) if they have no way to actully try my stuff?

Also in past, I have run very old binaries and they HAVE worked.

Indeed. The foundation (Linux kernel, GlibC, libstdc++, Xlib) is rock-solid. But everything on top is unusable mess.

People just don't do that, when a distro is built, they want a lean mean package optimised and targetting their current release.

Right. And they don't think about the needs of developers and as a result it's hard to develop anything usable for Linux if your program needs anything except the bare necessities (basically if you need to use GUI in any shape or form you are SOL).

Having been in position of doing ports, build and distribution for an ISV, I certainly DID not need 100% compatability. When you have high value software, for decent money, then you can invest in glue at install time.

Of course. But wast majority of the software is not high value software which you develop for decent money but more like “two guys in garage who have saved enough to dedicate half-year of their life to some problem”. They don't have the luxury to chase differences in the distributions and when their savings run over they release their software and switch to something else. They don't ever touch it again. Well, may be they will do a bugfix if it's broken by a new Windows (Android, MacOS, Linux) version, but then again, may be not.

Just look how difficult it is for MS to move userbase forward from XP with Win8, despite being in a position to dictate to OEMs and end users.

Sure. But that happens because they actually care about users. When users refuse to budge they try to convince them, try to offer some incentives, try to do some redesign, etc. Only when they are sure that only tiny percentage of users are unsatisfied (mostly non-payong ones in China) they abandon the old version. Compare with “if you don't want to play by our rules and upgrade two times per year the you could go away” approach of many distributions. Sure, people do go away… to MacOS or Windows. These OSes may be awful “vendor-controlled” ones, but their creators actually hear user's complains—which is not something you could say about Linux distributions.

Also, even if a distro goes out of it's way to support alien software, the OS/2 story tells us where that leads, ISVs not targetting native.

Sure. But eventually this becomes your only option till you die off comepletely. I applaud optimists like rqosa who believes that when you are thrown under the bus you could survive by using thick enough suit, but I somehow don't see such feature.