LWN: Comments on "musl libc 1.0.0 released"

I hope the licence changes

bronson — Thu, 10 Apr 2014 18:09:00 +0000

Like I said: in one case the mfr gives it to you, in the other case you need to steal them.

If you still can't see a difference, consider trying to sell the product you integrated.

I hope the licence changes

khim — Wed, 09 Apr 2014 22:38:40 +0000

Oh come on, there's a big difference. In one case the manufacturer is giving you the keys and crappy integration kit with the phone (or the line of phones you're going to sell). In the other case (xbox, wii, iphone) you are stealing the keys and assembling a crappy integration kit on your own from shady forum posts.

So. What's the promised “big difference”? What's the difference between Free60's or gc-linux's “integration kit” and some questionable Windows-only binaries offered by Chinese manufacturer?

P.S. Note: I'm not talking about iPhone here. Early versions of iPhone may be somewhat close to XBox360 or Wii WRT in the availability of “integration kits” department, but recent models can not be completely hacked like your could do with XBox360 or Wii. Perhaps you've meant these early models? They are “completely tinkerable” by these standards, too: you can even install Android on them!

I hope the licence changes

bronson — Wed, 09 Apr 2014 21:24:03 +0000

Oh come on, there's a big difference. In one case the manufacturer is giving you the keys and crappy integration kit with the phone (or the line of phones you're going to sell). In the other case (xbox, wii, iphone) you are stealing the keys and assembling a crappy integration kit on your own from shady forum posts.

Yes, it's true that a lot of these phones have awful and incomplete SDKs, but that's mostly because they're shoved out the door with the same haste as the hardware and software.

I hope the licence changes

mikachu — Wed, 09 Apr 2014 17:33:27 +0000

Android is a funny example, considering Google is moving functionality out of the open source parts into the 'google framework' closed source part continuously?

I hope the licence changes

nix — Wed, 09 Apr 2014 17:19:00 +0000

You said, about GCC:

the external project was recaptured via political maneuvering and a packed "steering committee" that gradually took power away from the fork's maintainers

This crazy conspiracy theory bears no resemblance to anything I experienced or anything I have ever heard. The egcs fork's maintainers were 'everyone who worked on GCC', including to some degree Richard Kenner, the then maintainer of the original project, though he was sort of snowed under trying to do bidirectional merges, mostly from egcs to GCC.

The allegedly 'packed' steering committee explicitly has no technical role so cannot 'take power' in any sense that I can imagine, and its members have always been major GCC contributors in any case and never people parachuted in by outside bodies. I can't think of any member of the SC about whom I could have thought 'hey, that guy shouldn't be there' though in a few cases I was left wondering why that guy wasn't appointed to the SC years before :) in most cases it was probably due to his then employer. Just as you would not expect of a body designed to be packed with "suitable people", no one company can employ a majority of SC members.

I hope the licence changes

khim — Wed, 09 Apr 2014 16:31:30 +0000

If you want to call these “tinkerable” then basically everything out there is “tinketable”—except for a few devices which include things like Secure Boot.

These phones come with POS which is called “source code” but which includes huge amount of binary blobs and often lacks important functionality. If it comes with source at all. Sure, you can grab blobs for other such similar phone and often they'll work (because their creators were just too cheap to afford custom hardware), but in the end it's huge mess which it very hard to deal with.

P.S. By that measure XBox360 and Wii are “perfectly tinkerable devices”: there are plenty of guides over there which explain how you can grab bunch of blobs from different places, tie them together and flash the resulting image.

I hope the licence changes

bronson — Tue, 08 Apr 2014 23:41:22 +0000

China is doing a decent job of making cheap, tinkerable phones. Alas, they're gray market and not being exported (not even on Ali/ebay).

The west may catch up... Time will tell.

I hope the licence changes

mjg59 — Tue, 08 Apr 2014 18:34:47 +0000

"Those who can tinker with phones tend to have higher than usual income."

Well, sure - and it'll stay that way if the only tinkerable phones are expensive ones.

"Also, previous generations of hackable phones can be bought for a reasonable price."

It's easier to budget for a subsidised low-end handset than it is to save enough to buy a phone in one go, even if the overall cost ends up being lower.

I hope the licence changes

Cyberax — Tue, 08 Apr 2014 18:15:21 +0000

Those who can tinker with phones tend to have higher than usual income. Also, previous generations of hackable phones can be bought for a reasonable price.

And if we're not talking about phones, then we have projects like RaspberryPI which are dirt-cheap.

I hope the licence changes

mjg59 — Tue, 08 Apr 2014 17:52:03 +0000

Hackable alternatives tend to be more expensive. It's unfortunate if the freedom to tinker ends up only in the hands of those with enough money to buy high-end hardware.

I hope the licence changes

bronson — Tue, 08 Apr 2014 17:37:45 +0000

That's true. But, if you're talking about phones or access points, the locked down devices tend to suck anyway.

The market has been doing a good job of ensuring hackable alternatives exist. And that's all that can be expected, isn't it? It would be unreasonable to have licenses or the law eradicate lock-in altogether.

(I've got to admit, the quantity of geeks using iPhones shocks me. You'd think they'd be a little more wary. But, of course, there's nothing the GPL or FSF can do about this.)

I hope the licence changes

mjg59 — Mon, 07 Apr 2014 19:20:56 +0000

"the FSF's great fear of Tivoization still hasn't happened"

Really? There are millions of shipping Linux devices where the only way for users to make use of their freedoms is to exploit bugs in the bootloaders - in some cases users are only able to modify the kernel because they can kexec from the shipped one! For millions more, even that's impossible.

I hope the licence changes

bronson — Mon, 07 Apr 2014 19:05:43 +0000

Dunno, ponderous licenses come with a heavy cost. The industry doesn't seem to think it's worth it these days (I sure don't). 2-clause BSD, MIT, and OSI projects are thriving, and yet the FSF's great fear of Tivoization still hasn't happened.

If things start going back to the way they were in the "open" 90s then, yes, I'll get worried. Maybe then I'll try to understand every corner of the GPLv3. But, for now, I love the freedom of programming and I love contributing to projects without having to think about all the implications.

I guess in 20 years we'll look back and see which position was more justified: my optimism or your pessimism. :)

I hope the licence changes

landley — Fri, 28 Mar 2014 14:22:28 +0000

> I'm not going to reply to each part,

Yes, I've noticed. I'm not sure if it's "any point you can't refute is ignored" or "any point you can't understand is ignored", and I'm not sure who of the two qualifies as giving you the benefit of the doubt.

I wasn't going to reply to any of it (which I'm sure is your definition of victory), but wound up reading again when the weekly edition included this article, and I'm curious about one thing:

> Your point #1 had factual errors, as I pointed out.

The original "point #1" reply from you was:

> > Linux isn't open source because of the FSF
>
> Linus moved from a non-free licence to the GPL because
> he liked GCC and its licence.

To which I pointed out that Linux 0.0.1 was already open source. Switching to the GPL removed the "no commercial use" clause, but Linux was already open source. The earliest versions were only distributed _as_ source.

You seem to consider me describing Linux 0.0.1 as open source as a "factual error". I don't see how.

As for Linus liking stuff, he preemptively rejected GPLv3 in 2000 ala http://lkml.iu.edu//hypermail/linux/kernel/0009.1/0096.html and he was already publicly saying "I don't trust the FSF." by 2001 ala http://www.win.tue.nl/~aeb/linux/lk/COPYING-modules.txt although I'm told his real falling out with the guy was when he first met him at the ~1996 O'Reilley conference, according to one of the other panelists at that conference. But throwing references at you gets us back into the uselessness of trying to reason someone out of a position they didn't arrive at by reason.

And as long as I'm here (http://xkcd.com/386), to avoid being hypocritical about "addressing points raised", your #3 and #4 are pure opinion but in #2 you say:

> Even FSF has left glibc as lgpl 2.1+ rather than v3. But you'd never
> mention that, because it doesn't help you argue that the world's
> problems are RMS's fault.

Allow me to quote Ulrich Drepper, maintainer of glibc until 2012:

http://lists.altlinux.org/pipermail/devel/2001-August/003...

> Stallman recently tried what I would call a hostile takeover of the
> glibc development. He tried to conspire behind my back and persuade
> the other main developers to take control so that in the end he is in
> control and can dictate whatever pleases him. This attempt failed but
> he kept on pressuring people everywhere and it got really ugly.

Ulrich may have been a bit of a bastard, but he was right that as soon as Roland McGrath took over his glibc 2.5.1 release announcement (back in 2007) said:

http://permalink.gmane.org/gmane.comp.lib.glibc.announce/3

> The 2.5 and 2.5.1 releases of the GNU C Library are licensed under the
> LGPL version 2.1, and GPL version 2.1 for the non-library programs
> included, or any later version. This means you are free to redistribute
> this version under the terms of the LGPL version 3 or the GPL 3, as well
> under as the version 2.1 terms of the 2.5 release. We expect that the
> next non-bug-fix release of the C Library will migrate to a newer LGPL
> version.

So yes, representatives of the FSF announced that glibc would move to LGPLv3. Of course Red Hat (which funds most glibc development, ever since it bought Cygnus) stepped in and went "er, no it won't". But it's not still v2 because Stallman didn't _want_ it to v3, it's still v2 for the same reason the kernel is: because the FSF didn't have the political leverage to enforce its views over code it didn't produce.

(And before you object to "didn't produce", the glibc 2.0 line was Ulrich Drepper's fork from the hurd libc, aimed at adding thread support. Ulrich was quite clear contained none of the original gnu code after a few years, the same way BSD didn't have any AT&T code left in it. When the Linux developers decided threading was a thing they should do, replaced their own libc5 with Ulrich's "libc6" (which the FSF named glibc 2.0 about the same way EGCS became glibc 3.0; the external project was recaptured via political maneuvering and a packed "steering committee" that gradually took power away from the fork's maintainers). Did I mention computer history is a hobby of mine? Track this stuff down and interviewing the people who actually did it is my idea of fun.)

Of course glibc and the sustained kernel campaign (resulting in the "leave britney alone" document at https://lwn.net/Articles/200422/) weren't the only attempt to shove GPLv3 down obviously unwilling people's throats. Did you notice the way that the last GPLv2 version of Binutils, release 2.17 from 2006, isn't on the FSF's website anymore? It was quietly replaced in 2011 by "binutils-2.17a" with a symlink from the old version to the new version. The only difference between them: the new one had GPLv3 source files added to it. So people trying to stick with the last GPLv2 better be paying attention and have the old one cached locally (as I was and do):

ftp://ftp.gnu.org/gnu/binutils

Or the way GNU's Savannah (sourceforge clone site) declared that GPLv2 was not a "free" license and they wouldn't host GPLv2 only projects anymore:

https://lwn.net/Articles/176582/

To be clear: a BSD license was still fine, but GPLv2 was not. Because the FSF said so.

The FSF's definition of "freedom" does not include "choice". (They keep using that word. I do not think it means what they think it means.)

All that said, I tried not to phrase any of the above "as a question, because I don't care about your answer anymore. Go argue with http://lucumr.pocoo.org/2013/7/23/licensing/ or somebody, I'm out.

Rob

I hope the licence changes

khim — Thu, 27 Mar 2014 08:55:44 +0000

Gold does not support few esoteric flags supported by ld. These are mostly irrelevant for most of the packages.

I hope the licence changes

mathstuf — Thu, 27 Mar 2014 02:57:22 +0000

Good to know. I'll swap it out in my WebKit build and see how it fares next time I have to build it.

I hope the licence changes

flussence — Wed, 26 Mar 2014 22:45:56 +0000

This was done on a Gentoo system, where all of that is hidden behind a "binutils-config --linker ld.gold" command. That's just a bash script which boils down to doing one (very paranoid-and-careful) link like you describe. The actual ld binaries are hidden in a symlink forest, at least on this distro, so watch out.

I remember attempting to use gold causing scary breakage when it was first introduced a few years ago, but it looks like they've ironed out those bugs now and I've been using it for my home systems without any incident for a few months.

I hope the licence changes

mathstuf — Wed, 26 Mar 2014 20:52:33 +0000

Is it just as easy as changing the linker from /usr/bin/ld to /usr/bin/ld.gold? What gotchas are there?

I hope the licence changes

flussence — Wed, 26 Mar 2014 17:53:03 +0000

I switched from vanilla ld to gold a while back, and the improvement when building recent versions of Firefox/Webkit on a low-RAM machine is significant (in my case, it makes all the difference between having a working web browser in the morning or seeing 8+ hours of compiling end in a linker OOM).

I hope the licence changes

mathstuf — Wed, 26 Mar 2014 17:46:07 +0000

> This in turn motivated lots of permissibly-licensed projects like Clang and LLVM.

Shouldn't this read "This in turn motivated lots of *new development interest in* permissibly-licensed projects like Clang and LLVM."? LLVM is older than the GPL3 I thought.

I hope the licence changes

Cyberax — Wed, 26 Mar 2014 17:35:38 +0000

> For non-free versions of the Apache webserver, I don't have examples. I just heard recently that they are common and that the Apache devs thought they could prevent it by frequently changing the ABI but they see now that it hasn't prevented the phenomenon.

Exactly wrong. Apache has an ABI stability promise for its modules and generally tries NOT to break them.

And while there are closed-source modules and even complete forks of Apache, they mostly serve as a useful halo, not as a distracting force. I.e. a company which makes, say, a PHP accelerator module is more likely to contribute fixes and improvements to Apache core instead of keeping them in-house.

> So we've one saying it ruined everything, and another saying no one uses it. If a concrete problem really existed, it would be clearly one or the other.
The are not mutually exclusive. GPLv3 did damage mostly to vendors' expectations. This in turn motivated lots of permissibly-licensed projects like Clang and LLVM.

So now GPLv3 is simply becoming irrelevant. Apart from Samba, I can't even remember a GPLv3 project that is irreplaceable in my software stack.

I hope the licence changes

coriordan — Wed, 26 Mar 2014 17:10:07 +0000

For non-free versions of the Apache webserver, I don't have examples. I just heard recently that they are common and that the Apache devs thought they could prevent it by frequently changing the ABI but they see now that it hasn't prevented the phenomenon. I hope what I heard is wrong :-) and that fully free versions are prevailing.

> Nobody uses GPLv3 so it's not a problem.

So we've one saying it ruined everything, and another saying no one uses it. If a concrete problem really existed, it would be clearly one or the other.

I hope the licence changes

Wol — Wed, 26 Mar 2014 15:51:54 +0000

You can't swap the means and the end over, because one has to be justified in terms of the other.

If you justify writing Free Software in terms of support, you get a profit. You sell your support services, your staff write Free Software to provide that support. You hire the people who write Free Software in order to enhance your reputation and ability to provide support.

Google does the same sort of thing. Each employee spends 10% of their week working on a personal project, most of which I believe end up in the free software ecosystem. But that buys Google staff morale, it earns them good Karma, and it attracts talented guys to work at Google. Google's end is to sell advertising, writing free software is both a means towards that, and a defence against Microsoft.

If we swapped the two over, with the *end* being to write Free Software - mmm - I think I see where you're coming from ... "we've got to provide support to bring in the money to write Free Software". Red Hat are pretty much the only company in that space so we can't really say which way round it is. But I still think that if two companies came up against each other, the company that viewed Free Software as its *means* would be the one that came out on top. It biases your decisions, such that the company that viewed Free Software as the *end* would probably make less money, and would inevitably sink.

Cheers,
Wol

I hope the licence changes

raven667 — Tue, 25 Mar 2014 19:44:48 +0000

I don't see any logical reason why you couldn't reverse the two in your example, RH support and services and the means to the end of more Libre Software, especially since RH invests in many resources in areas where they don't compete with support services.

I hope the licence changes

Wol — Tue, 25 Mar 2014 18:43:29 +0000

> If doing so would help Google's business

Thank you for proving my point. That one sentence says it all!

If it "helps Google's business" then it is a means, not an end. And as I said, Google sees "Gratis Source" as a very effective means.

Any company that views Libre Software as an *end*, on the other hand, is going to go bust. Writing software is expensive. Without income, the result will be inevitable ... (and no, Red Hat is NOT a counter example - Red Hat's *end* is support and services, "Free Software" is merely the *means* to that end).

Cheers,
Wol

I hope the licence changes

raven667 — Tue, 25 Mar 2014 16:03:26 +0000

Look also at the recent example of Tim Cook and Apple building solar panels at their data center, where he told a 1% shareholder to get bent when asked to prioritize ROI over the environment, in the absence of any regulation requiring them to care about the environment. "Fiduciary Responsibility" is a joke, and not a funny one.

I hope the licence changes

jwakely — Tue, 25 Mar 2014 11:58:09 +0000

> Plus, as a commercial business, it's probably illegal for Google to have "increasing the amount of Libre Software" as one of their *ends*!

Illegal my shiny metal ass.

If doing so would help Google's business then it would be in their stakeholders' interests and it is certainly not illegal to run a business in its stakeholders' interests.

I hope the licence changes

Wol — Mon, 24 Mar 2014 21:00:24 +0000

Why oh why are you so vocal about trying to force others to share your motives? Why oh why do you *delude* yourself that others share the same aims as you? Why oh why are you unable to see the difference between the *means* and the *end*???

At the end of the day, it's obvious that your *END* is to maximise the amount of Libre Software out there. Okay, that's a highly laudable goal but the harsh reality is that to the majority of people out there, that is a total waste of time. To be brutally honest, the majority of people out there are incapable of taking advantage of it. The aim of the GPL is to give freedom to the *END* *USER*, most of whom are incapable even of understanding what they are being given.

On the other hand, people like those at Google have decided that Gratis Source is a highly effective *means* to their ends, whatever they are (and as far as Google is concerned, it seems that *end* is to "survive whatever MS can throw at them, and come out on top rather than going bust"). Plus, as a commercial business, it's probably illegal for Google to have "increasing the amount of Libre Software" as one of their *ends*! (Hence the ban on (L)GPL software - the source promptly ceases to become gratis, so that restriction is NOT silly at all from Google's p-o-v. ALWAYS assume that, if something seems silly to you, it is because you are lacking information. And in this case it seems you are ?willfully? blind to Google's real motives.)

I certainly don't have "increasing the amount of Libre Software" on my list of ends. I'm very much an advocate of the Pick/MV database model, and I will happily write Libre Software if I see it advancing my ends (which I do), but my motive most definitely is not Libre Software.

The more you tell people they have to align their motives with yours, the more you will piss them off. What you are very successfully FAILING to achieve, is to persuade others that they will achieve their ends more easily, by helping you to reach yours. If you want Libre Software as an *END*, you need to show others that Libre Software is their *MEANS* to their end. Demanding that they change their end to match yours is a recipe for failure, and can even easily make enemies and HINDER YOU in achieving YOUR end.

Cheers,
Wol

I hope the licence changes

mm7323 — Mon, 24 Mar 2014 19:56:47 +0000

While permitted, it doesn't really work very well in practice. In trying to release software and installers in binary forms for different distributions, I've found it quickly becomes a mess as minor differences in library version, location and naming mean a package or installer needs to be aware of not only the distro being targeted, but often the version too. Other system differences just make the whole distro landscape a mess.

I welcome musl if it means being able to statically link against a core library without any further concern and produce a program which is easy to install and start without having to consider 10 different platforms.

Otherwise targeting only popular and stable distros (e.g. RHEL and derivatives) is the other somewhat limiting option.

I hope the licence changes

landley — Mon, 24 Mar 2014 02:29:19 +0000

Yes and no. 1) The same claims were made about filesystem compression back in the Stacker days, didn't really pan out. 2) Latency is a thing. 3) Battery life is a thing.

I understand compressed swap, and I understand compressing pages as an alternative to swap. But it's not the same as actually having more memory, and adds a lot of uncertainty to resource availability that's going to make performance tuning way more complicated.

(Why would RAM manufacturers worry about declining PC sales? My phone has 2 gigabytes of ram. Flash is solid state storage. They've got plenty of market opportunity right now. :)

I hope the licence changes

landley — Mon, 24 Mar 2014 02:21:54 +0000

Current data on system lifetime/replacement, albeit from the mac side of things and focusing on the new stuff: http://tech.fortune.cnn.com/2014/03/21/apple-ipad-cirp-re...

Says macs are cycled out every 2-4 years, which is similar to the ~3 from PCs in 2006.

Phones are cycling faster, but they're A) heavily subsidized, B) new technology still maturing. (Well, still doing the whole disruptive technology upward attack thing to consume the established market's revenue.)

I hope the licence changes

Cyberax — Mon, 24 Mar 2014 01:52:28 +0000

> Your point #1 had factual errors, as I pointed out. And while, yes, Apache does use a push-over licence, many users get no, or reduced freedom because they're using non-free forks or non-free modules.

Are there any major non-free forks of big Apache-licensed products that added major new functions? I can't really remember any.

>For point #2, I think you're exaggerating the incompatibility problems caused by GPLv2 vs V3.
Duh. Nobody uses GPLv3 so it's not a problem.

I hope the licence changes

ibukanov — Sun, 23 Mar 2014 20:15:10 +0000

Hm, I have seen the xz compressor routinely archives the factor of 3 and more for binaries with/without debug information. Now, xz is not something that one wants for the compressed RAM, but then LZO compression as used by zswap or zram gets the factor about 2 for binaries. So while 4GB is not enough, 8GB of physical RAM should be enough for anything that maxes at about 14GB. And 8GB limit gives significantly more options in notebook selection...

I hope the licence changes

mathstuf — Sun, 23 Mar 2014 15:06:45 +0000

If you can compress the final link (which takes 8-10G depending on debug symbols) to 4G, you'd have an algorithm I'd like applied to the binaries in packages ;) .

I hope the licence changes

ibukanov — Sun, 23 Mar 2014 14:38:00 +0000

Thanks for writes and historical links you have posted. I would love lwn.net just for such comments :)

I hope the licence changes

ibukanov — Sun, 23 Mar 2014 13:56:16 +0000

No need to upgrade - just enable one or another form of a compressed RAM. I guess RAM manufacturers hate that development even more than declining PC sales.

I hope the licence changes

coriordan — Sun, 23 Mar 2014 11:00:16 +0000

I'm not going to reply to each part, but you throw in things like the licence change being "a reasoned decision", as if that refutes my comments. I never said the licence was an accident.

Your point #1 had factual errors, as I pointed out. And while, yes, Apache does use a push-over licence, many users get no, or reduced freedom because they're using non-free forks or non-free modules.

For point #2, I think you're exaggerating the incompatibility problems caused by GPLv2 vs V3. They're quite rare and most can be solved by making a v2 exception for the relevant files. Even FSF has left glibc as lgpl 2.1+ rather than v3. But you'd never mention that, because it doesn't help you argue that the world's problems are RMS's fault.

For point #3, so?

For point #4, yes, and? RMS didn't write the (silly) Android trademark licence, nor does he control what types of computers sell most. Your mish-mash of points don't back up your criticisms of RMS, and they don't support the idea that GNU should give up on using their licences to create an incentive for developers to release their software as free software.

I hope the licence changes

Wol — Sun, 23 Mar 2014 01:23:08 +0000

And the GPL/LGPL stuff is pretty irrelevant in certain quarters, too ...

In this system I keep intending to write :-) I'm not at all sure what licence to use. Because I want it to be part of LibreOffice it'll almost certainly be MPL, which I'm perfectly happy with. BUT.

Because of the practicalities of the sort of system it is, I'm of the reasonably informed opinion that the MPL, GPL and LGPL are all pretty much FUNCTIONALLY EQUIVALENT! Or, to word it slightly differently, the clauses that differentiate these licences will be legally void in these particular circumstances!

Basically because large chunks of the system will be necessarily shipped as source, and even when they aren't stuff will generally be shipped as small stand-alone modules.

Cheers,
Wol

I hope the licence changes

landley — Sat, 22 Mar 2014 23:34:45 +0000

> How much of this is due to the need for new computers dropping as well?

"640k ought to be enough for anybody?" :)

A few years ago I co-authored a paper with Eric Raymond on the upcoming transition from 32-bit to 64-bit PC hardware. The table of historical computer memory sizes is at:

http://catb.org/esr/writings/world-domination/world-domin...

The big mistake I made in that analysis was missing the stretching of the active window in the 1990's due to the internet. (The old "low end is 1/4 the high end" rule broke when 56k modems became the main system bottleneck; this was the market vacuum that launched AMD's Athlon and which intel shipped the Celeron to plug, an excellent article about which was http://www.forbes.com/forbes/1999/0125/6302088a.html).

The current low end is about 1/16th the high end because that stretched the range. That's why Microsoft could recover from the vista debacle and ship a 64-bit capable Windows 7 in time to retain its market dominance, the low end migrated ~3 years after we though it would.

But even Vista didn't depress the PC market anything like we see now. Moore's Law continued unabated, price/performance ratio improved 50% every 18 months and if it wasn't improved performance it was lower price. Dell drove Compaq to break the magic $1000 barrier, then the $500 barrier went down a few years later, and now we have $250 netbooks. But all that was _before_ the current decline.

The above is a long way of saying "we looked into this, and we're pretty sure the drop in PC sales is due to phone and tablets replacing them, not due to anything internal about the PC market".

> It took me 3 years to upgrade my last one (and the old parts are now
> running a server) and I don't see a new one in the next half dozen
> years or so (modulo disk space…which on mobile and tablets typically
> would require a new device when hitting the limit).

That's actually about average. PC replacement time is a bell curve with a peak around 3 years and the little dotted vertical bars at a year and a half and 7 years. I forget what the dotted bars mean but it's basicaly "the normal range". (Alas, I don't remember where I saw that. Probably some random research material when I wrote the above paper. It wasn't an online reference I could snag a URL to or I'd still have it. Maybe a library book?)

> There are a lot of confounding factors here that need to be split
> out before placing the blame of the PC/laptop decline on smaller
> devices.

There are professional research people whose job is to work that out. I used to just bug Stacey Quandt for info, but we fell out of touch a year or two before she married Ted T'so. These days I mostly just follow links in news articles to their source data. Gartner and such charge giant piles of money for their research reports, so I often have to go by secondary sources there, but you can meet up with the people who actually _make_ this stuff at things like CELF (now "The Linux Foundation Embedded Linux Conference by The Linux Foundation"), and it's their _job_ to know why people do or don't buy their stuff. And they'll usually talk in person, off the record. :)

tl;dr "Nope, we looked, pretty sure it's phones/tablets."

Rob

I hope the licence changes

landley — Sat, 22 Mar 2014 23:02:13 +0000

> (Most of your points have no relation to my comment, so I don't know
> what to reply to...)

So saying "Stallman is out of touch, your argument is based on a false premise, copyleft may no longer be viable, here's why" in the previous message has no relation to your "why isn't the project using a copyleft license?"

Early in its development musl was LGPL, just like glibc and eglibc and uclibc and dietlibc and so on. The maintainer (back when it was still almost entirely his code) changed the license on his copyrighted code (and thus with a few minor clearances the entire project) from LGPL _to_ MIT. Doing so was a reasoned decision on his part.

I thought my post directly addressed your concerns, by explaining why considering non-copyleft licensing actually superior to copyleft licensing, for the purpose of promoting the use of open source software and "software freedom" in general, is at least why it's a viable intellectual position to hold.

You seem to be saying you don't care to understand why he did it, or why someone might ever evaluate both positions and find copyleft wanting. Maintaining willful ignorance of opposing viewpoints seems unlikely to make you a more effective advocate.

Back in 2005 I was an enormous fan of the GPL. I was the _pro_ GPL side of this panel:

http://sf.geekitude.com/content/pros-and-cons-gnu-general...

And I launched the world's first actual GPL enforcement suits that went to trial and resulted in a legal verdict:

http://www.softwarefreedom.org/news/2007/sep/20/busybox/

I didn't leave the GPL, the GPL left me. I had _reasons_ for changing my position, and explained them at some length in a talk I gave at Ohio LinuxFest last fall. It was called "the rise and fall of copyleft", and there's an MP3 of it at:
https://archive.org/download/OhioLinuxfest2013/24-Rob_Lan...

(I need to do a better version of the talk, by the time I presented it I'd trimmed my research material down to 3 hours and had a 45 minute slot, and what's there involved pointing people at a series of web pages that aren't in the audio recording. Oh well...)

> Linus moved from a non-free licence to the GPL because he
> liked GCC and its licence.

No, Linus moved from a "no commercial use" license to the GPL. Linux was always freely downloadable source releases you could modify and redistribute to your heart's content. (Have you read Linus's biography "Just For Fun?" I have, and over a dozen interviews with him.) The primary result of Linus moving to the GPL was _allowing_ commercial use, and these days non-commercial development of Linux has been declining. Remember last week's kernel status report, which said "the number of contributions from volunteers is back to its long-term decline"?

https://lwn.net/Articles/590354/

That's _vanilla_ Linux. Something like 90% of all Linux deployments are actually Android (b as in billion), which develops the code in house and then does periodic abandonware drops they never accept changes back from. (By the time you see it outside of Google, it's ~6 months out of date.)

The GPL is almost orthogonal to both the success and the openness of projects. QT used the GPL to sell commercial licenses (try the GPL version, then get a proprietary license for _real_ use), which prevented them from accepting any external contributions to the project. Most flavors of BSD are _each_ more actively developed and widely deployed than The Hurd without even opening the "Darwin" can of worms.

>> Creative Commons and Wikipedia are not related to (or inspired by)
>> the FSF....
> Ha! Read your history. Lessig and Jimbo massive RMS fans and are
> very vocal about it.

Actually, computer history is a hobby of mine:

http://landley.net/history/mirror

If you've ever read the book "Hackers" by Steven Levy (which devotes a chapter to Stallman's early years), the Ken Olsen interview in that mirror is basically the other half of that story (where did the TX-0 come from, and so on). Probably the bit of that page I'm most proud of was tracking down interviews with all four of the Intel 4004 principals (Ted Hoff who actually designed it of course, Gordon Moore the cofounder of intel who was his direct boss at the time, Federico Faggin (the layout engineer who actually produced the hardware, and who went on to found Zilog), and Masatoshi Shima (of Busicom) who was the actual customer who ordered the 4004 in the first place. I like to be thorough, and understand things from all angles. (The 4004 was the first microprocessor, FYI. Kind of a big deal.)

I first emailed Richard Stallman in 1998 (the apology I posted to linux-kernel a few years later got reposted in places like http://www.kuro5hin.org/?op=displaystory;sid=2002/8/1/045...). I drove to Boston to interview Richard Stallman in February 2001 for the above computer history research, spent a couple hours talking to him and even let him borrow my car to run an errand (he left it unlocked). Of course I've read his biography "Free As In Freedom", and have a copy mirrored on my website: http://www.landley.net/history/mirror/faif/toc.html

When I say "Stallman is hopelessly out of touch", I mean things like him signing his emails 'Richard Stallman, Principal developer of the operating system often inaccurately called "Linux"' (Yes really, see http://www.theregister.co.uk/2001/04/15/richard_stallman_...). _I_ am not the one speaking out of ignorance here. I can give you extensive citations. :)

If you haven't got time to listen to my Ohio LinuxFest talk (or the recent Linux Luddites podcast that interviewed me on some of the same topics last week at http://linuxluddites.com/shows/episode-11/), you could read the two computer history blog entries I wrote in 2010. The first is "history of open source unix", the second is explicitly about the FSF:

http://landley.net/notes-2010.html#17-07-2010
http://landley.net/notes-2010.html#19-07-2010

You don't have to agree with any of this, but my point is that it is possible for people to disagree with you for legitimate reasons.

> > I might have done a little bit of research on the subject. :)
>
> Feel free to do some more :-p

When I don't know things I _ask_ people. For example, I drove to California to interview Peter Salus who was both Vice President of the Free Software Foundation and Executive Director of Usenix. (Turned out he lived about 2 miles away from me back in Austin, and we became friends, but the point was he answered my questions. I've since organized two Linux conferences, Penguicon and Linucon, that he was a guest at. He's written several unix history books, the two of his I own are "A Quarter Century of Unix" and "The Daemon, The Gnu and the Penguin".)

You are stating a position, and when it is factually challenged you either repeat your initial assertions or claim anything but direct agreement is off-topic. That's not advocacy, that's evangelism. As the above Linucon panel where I was pro-GPL shows, I have previously _changed_my_mind_, as has the musl maintainer (who posted to this thread as "dalias"). That means it should be possible for you to argue either of us back around, but the way you're going about it is not very effective.

Rob