Richard Stallman is responsible for the shrinking economy

You *have* posted this already in reply to Walter. You would be well advised if you see a physician - failing memory is something one should not take lightly.

Dimiter

Original message:

They may not need to make direct contact if they follow the GPC license. If you are talking about other cases, your comment may be more projection than reality. I can't say, of course. Only you can. But it crosses my mind.

In any case, I know I would certainly make appropriate contact. So perhaps ignorant of decidedly comprehensive knowledge, its arguably better to simply speak for yourself rather than pretend to speak for so many others.

Jon

He just made some sort of silly bit with someone about what would happen if he contravened Usenet conventions in a reply. The result has been at least 6 useless messages cluttering the group.

--
 [mail]: Chuck F (cbfalconer at maineline dot net) 
 [page]: 
            Try the download section.

I post it in response to evidence that it is needed - i.e. to top-posting.

--
 [mail]: Chuck F (cbfalconer at maineline dot net) 
 [page]: 
            Try the download section.

... snip ...

Why do you say that? I concede that it doesn't adhere to ISO by default, but the command:

gcc -W -Wall -ansi -pedantic ...

runs it as a C90 compiler. Replacing ansi by -std=C99 makes it a C99 compiler (almost). Modulo bugs and library.

I don't know about the test suite.

--
 [mail]: Chuck F (cbfalconer at maineline dot net) 
 [page]: 
            Try the download section.

The open source has had the effect of putting price pressure on innovation in software technology by making lowering the potential reward ./ risk ratio. The open source movement has not embraced standards and has not generally participated in the standards process.

GCC has not participated in any meaningful way in WG14 the ISO international standards group (represented in the US by ANSI). What is more disturbing is GCC has not made a significant attempt to be ANSI/ISO compliant. The test suite distributed with GCC is a regression list of past bugs and development \test cases not a language syntax test of organized code generation test.

This thread has generally shown respect for IP rights and licenses the exception has been respect for standards. C standards organizations are partly paid for by publications to users. (Standards participants are not paid) It is disturbing that some of the strongest advocates for open source are also willing to violate standards copyrights and undermine the organizations that help everyone open or otherwise.

It is a question of degree. Academic innovation based on the GCC or other FOSS core has been limited. There are a lot of major University projects that just have not moved the technology forward in any significant degree.

Regards,

-- Walter Banks Byte Craft Limited

I assume here you are speaking from your own experiences with the c standards processes.

Since you probably have some ideas about this, how would a single representative be selected and representative of the "open source movement" in the standards processes? I mean this seriously. I'm curious how that might be made to work well and I'd like to hear your thoughts about it.

It might be that the well-worn paths in the standards processes are tuned, more because of historical circumstances, to the traditional models that existed earlier and it may be expecting a lot to imagine a situation with an "open source movement" working without some significant adaptations.

In the above, I am intentionally conflating "open source movement" with GNU c. I know I'm doing that "on the table." I intend it as a prod to ask you to tease the two things apart, again. In so doing, that I understand the comments better.

Well, that regression list is for obvious reasons -- in the flurry of contributed activity unlike what a single organized group working together experiences, to "form a floor" beneath which they cannot (hopefully) sink. Clearly, it's needed.

What you are really talking about is the ANSI/ISO compliance and participation. I am already asking you above about how that participation might meaningfully take place -- in a room _full_ of people representing commercial interests, it might be interesting -- at a minimum. But if you feel it can work, I'm interested in hearing how.

As for compliance, I have to admit my own ignorance. Can you elaborate with some examples so that I can understand and comment?

Do you have any insider information or educated guesses about why it is that FSF hasn't participated, or other significant groups in the open software movement? Other than just to say that you are concerned they haven't participated more? Frankly, I've only some vague guesses about it and I fear most of them are rather ill-informed ones at that.

I'm curious not so much about the simple facts you claim as about why you think they have come to be that way, today?

I think the wider thesis you propose isn't about academic innovation. Taking your larger arguments here as a whole, you seem to me to be complaining that this lack of innovation across the board -- academic and commercial.

In the commercial spaces, I tend to agree. On the academic side, I don't as much. I have pulled down and had a chance to read a few from lots of very good papers coming out over the last decade and a half -- all of which date well after the (now) ancient 1986 version of the Dragon Book and JR Ellis' excellent Ph.D. thesis on Bulldog (some of which _could_ be applied well today, but isn't.) My limited exposure sees more of a lack of delivery, and less a lack of good research.

But on that lack of delivery, I might take your point that GNU c has had some impact in inhibiting risk taking. You'd know better than I would about that.

Another facet that crosses my mind is the sweeping change in those who consider themselves programmers in my lifetime. In my earliest days, you were pretty much a graduate of some kind -- often physics -- and this meant a very high level of caliber could be expected. And "consumers" were very large corporations that could afford the custom built, air conditioned rooms and the near-million-dollar expenditures for the hardware. Companies hired the best on all sides, and got it. Today, computing is accessible to nearly everyone. I've already commented before here about a student coming up to me, complaining that the 2nd year course seemed too hard and that maybe their choice to choose a CS degree instead of an accounting degree was wrong... But when, when CS degrees didn't exist and people got into computing from the physics or math departments, there was no such question in anyone's mind I ever met. Not on the radar scope. But today, we have almost anyone with almost any level of native talent becoming programmers here and there. Not bad. Not good. Just different. And the marketplace itself, because of that, is also different. And so are the relative levels of research for various areas, I suppose.

Maybe everything is just a two-edged sword. With choice and options and lower prices for consumers, there is a reduced level of innovation due to lack of excess profits to invest in research, for example. And a different consumer type, as well. A practical balance is probably the better we can hope for, if so. I wouldn't want the pendulum swung to one side or the other. In any case, there is no going back. Accountant types ARE choosing careers as programmers. It's life in the modern world.

Anyway, I'm curious why you think it is important GNU/FSF folks get involved in the committee activities and how they might meaningfully do so. I would have imagined you didn't care, before your comments, because the standards activities go on through thick and thin and with hardly a word or complaint. Now, I wonder why you think this is significant. Unless all this was really just about the impact on tool developers making enough to innovate and maybe where you conflate these two together because you see the close connections better than I do.

Jon

No? I'm pretty sure that that is one of their stated goals. I'm also pretty sure that if you submitted a bug report with a failing test case that demonstrates a non-compliance, then that bug would most likely get fixed, sooner rather than later. That's the way OSS works.

Standards participation generally involves expenses, either directly through standards-organization membership or indirectly through travel and time. Maybe some well-funded OSS contributors can manage that, but I'd guess that most can't, and it is certainly a broad church.

On the other hand (and in a corner that I'm more familiar with than GCC), there is the FreeBSD foundation, an organization established to collect contributions of money or equipment to the project, and distribute it as they think appropriate. This has enabled them, in the past, to buy into the Java testing and branding process with Sun. They also pay for (some) travel for conferences and the like. I think that GCC has a similar body, perhaps other projects do too.

I think that you'll find that OSS folk are, by and large, strong champions of open international standards: it's the only comfortable way that they can inter-operate with the rest of the world, as they are not generally able to sign NDAs. The other way is reverse engineering, which is tedious and unsatisfactory but has had some significant and notable wins in the past.

Cheers,

--
Andrew

Ah, I see. So you remembered you had posted these to Walter but you chose to be rude and repost them for the umptieth time.

Dimiter

As far as I can tell, GCC has provided a nontrivial amount of "existing practice" to standardize, and I also found people from WG14 meeting participant lists active on GCC mailing lists.

I have seen C99 features in GCC much earlier than in, for example, Microsoft or Borland compilers.

And of course they try to be ANSI/ISO compliant. But they also have a user base to serve. Just look at the vast amount of flamage generated when GCC implemented ISO aliasing rules. Plus, their user base wants to use GCC extensions, some of which were standardized by ISO under subtly different rules than long ago implemented in GCC. You cannot simply implement ISO 'inline' rules as long as you have a Linux kernel based upon GCC 'inline' rules.

I don't expect commercial compilers to be much different here. Microsoft will surely not make their compiler fully ISO compliant just to put a ISO sticker on the box, if that would mean MFC does not compile anymore.

So what? Both are just stochastic tests, but a regression test at least shows where problems were in the past (and _might_ thus be more likely to appear again).

Stefan

Jon,

Quick comments.

GCC at WG14. ISO representation is substantially an individual activity. GCC could be represented relatively easily. ISO standards are an intense activity. Ideas promoted at meetings need to be backed by well prepared arguments.

The push to retain C90 as the standard by many in "open source movement" suggests the opposite. C99 addressed a lot of serious language and library issues and to some extent projected the language forward. It addressed language issues of current practice, concerns raised by misra and established the difference between C and C++. Since the release of C99 several application specific TR's have been released further broadening the base of the language.

It was. (about academic innovation).

For example backend development projects (most recent LLVM) still don't address the failures of UNCOL.

There has been a lot of commercial compiler technology development that has advanced the art. Most unpublished. Surprisingly software patents would have put much of this material in the public domain by now if they had been made viable.

Regards,

-- Walter Banks Byte Craft Limited

Why do you think there is a "push to retain C90 as the standard" in open source? If you are talking about the standard to which most open source software is written, then C90 is the obvious target. Most open source projects are aimed for a very wide range of platforms, both modern and old, commercial and open source, and with a wide variety of compilers. The only sensible choice is C90 since C99 support is often patchy - there are very few compilers (open source or otherwise) that have near to full support for C99. The only other choice is to insist on gcc or reasonably compatible compilers (such as Intel's compiler, which implements many gcc extensions) so that the programmer can take advantage of language improvements in C99, and also gcc extensions. The Linux kernel, and gcc itself, are examples of such projects.

If you are talking about the compilers (mainly gcc, although there are other open source compilers such as llvm), I don't follow you. gcc has been implementing C99 since before C99 existed, and similarly with newer proposed standards and C++ standards. It certainly does not implement

*all* the features of C99 - few (if any) compiler/libraries do - but it implements the majority, especially those that are widely used. My understanding is that there are some language features that started life as gcc extensions, were then copied by other compilers, and finally made it into the standards or standards proposal. I don't follow the details of the standards well enough to give examples, except perhaps the use of "0b" for binary constants similar to "0x".

In the C++ world, a great many of the recent standards proposals have come from the boost project, which is a large open source project.

I can fully agree with you that more gcc (and perhaps llvm) participation in the C and C++ standards groups would be a good thing, but I disagree if you are saying that gcc is "living in the past" regarding standards - at least, in comparison to other compilers (it's a conservative business).

Would it be feasible to have a compatibility switch? I remember seeing this is earlier days to enable nested comments and $ in identifiers to be compatible with earlier versions.

--
Thad

FWIW, the UK allows private prosecutions, although I'm not aware of FAST (or BSA, etc) having brought any. Apart from anything else, it would be harder for them to win a criminal case due to the "beyond reasonable doubt" standard.

Usually, open source software aims at portability. As long as there are significant platforms that don't have a C99 compiler, people will recommend C90. Likewise, GNU coding standards have for a long time after C90 recommended to consider K&R C, because not all platforms had a C90 compiler.

The aim for portability is not unique to open source: I haven't seen commercial code fully exploiting C99 either. Okay, "//" comments, "long long", and declaration-statements are in use, but C99 is much more than that.

That last point is another argument for (partially) staying at C90: C90 is the least common denominator for library interfaces. It'll probably take a while until we see variably-modified types in library interfaces.

I wonder how something can "advance the art" if it is unpublished...?

Stefan

More significantly, it also allows them to change the licence in the event that any of the terms are found to be unenforcable, or are decided to be undesirable.

Changing the licence for a project with multiple contributors can be next to impossible. Even if the changes are uncontroversial, simply tracking down all of the contributors to obtain consent can be impractical.

I once modified the 68K GNU compiler to suit the register model of a complicated embedded application (heart of multi-million dollar ic-manifacturing machines.) Took me a couple of weeks. No, I don't charge as much as the average US lawyer.

They switched compilers, for the sole reason that the source was available. It saved their cash cow.

Groetjes Albert

--

--
Albert van der Horst, UTRECHT,THE NETHERLANDS
Economic growth -- like all pyramid schemes -- ultimately falters.
albert@spe&ar&c.xs4all.nl &=n http://home.hccnet.nl/a.w.m.van.der.horst

You are apparently against a free market. The reason quite a lot of people are in favour of a free market economy is competition. The advantage of competition that it drives the price down for consumers. It is not there for keeping the prices up for producers, especially those who don't deliver, those who can extort money from consumers only by holding governments hostage, or even bribing them. (Not that I'm in favour of a free market for just everything.)

GNU and free software are just playing in a competitive market and they *are* playing by the rules.

How is that different from any other cost Managers and Directors spend money on, except their own bonuses?

Groetjes Albert

--

--
Albert van der Horst, UTRECHT,THE NETHERLANDS
Economic growth -- like all pyramid schemes -- ultimately falters.
albert@spe&ar&c.xs4all.nl &=n http://home.hccnet.nl/a.w.m.van.der.horst

This is a very interesting statement. It actually means that in your opinion most programmers are doing work that is not economically essential. It is the task of a free market economy to get rid of that work. Are you in favour of an extortion/IP economy over a free market economy? Get used to it. Once humanity develops an idea, it is in the meme pool and stays there. Progress means there is less and less work to do in certain areas such as inventions and production. OTOH there is more and more work to do in education and taking care of the elderly. The present economic system is ill-equipped to take this on.

Remember, patents are there for allowing inventors to reap benefits from their invention, not for patenting trivial software such that it has to be reinvented and repatented in a slightly different form. The first is economically beneficial, the second is not. (And in the development of medicines, the same pattern is visible.)

Groetjes Albert

--

--
Albert van der Horst, UTRECHT,THE NETHERLANDS
Economic growth -- like all pyramid schemes -- ultimately falters.
albert@spe&ar&c.xs4all.nl &=n http://home.hccnet.nl/a.w.m.van.der.horst