Re: Does the LGPL encourage sharing of code?

Roger Fujii wrote:

>Anthony Taylor <[email protected]> wrote:
>
>>I'm *not* coming to Brett's defense here, as his statements are not very
>>intelligent (which is strange, because Brett isn't a stupid person, just
>>obsessive).  However, Brett is correct in that GCC's code isn't as
>>efficient as some compilers: for instance, Sun's compiler produces
>>better SparcV9 code, and IBM's compiler produces better PowerPC code.
>> ("Better" here means more efficient, faster code.)  Even Intel's
>>compiler produces better code than GCC.
>>
>
>plus they have better profiling abilities, better debugging and doesn't
>suffer version incompatibilies... the list goes on.  Forte produces better
>V7 code too (for the SPARC).  Heard the x86 version produces better code
>than the gcc too (but haven't verified this).
> 
>
I've seen it; it does.  Forte is a very nice compiler, and has an 
excellent development environment, as well.

>
>>All these compilers prove that GCC is *not* killing off other compilers.
>> Brett would be hard-pressed to come up with a *single* example of a
>>compiler which GCC has killed.
>>
>
>metaware.  watcom.   
>

I'm not sure about Metaware's case, but Watcom was killed by Borland and 
Microsoft, which was unfortunate; they had a great compiler.

>>In the PC world, Zortech was killed by
>>MS and Borland; Borland is slowly being displaced by an inferior
>>*commercial* product (MS Visual C/C++/etc).  In fact, every single
>>argument Brett has put forth is unsupported by the evidence.
>>
>
>Personally, I think the pro-GPL camp has less evidence shown for it here.  One can
>easily come up with simple economic reasons why *GPL has problems.  
>

I'd argue that economics is too complex to come up with rationales. 
 It's too easy to make any economic case make sense.  And economics as 
practiced today in the US is flawed; stocks are sold as commodity.  This 
is what allowed the "dotcom" (stupid name) economic phenomena at the end 
of the '90s, when any sane person could see there was no economic 
foundation for any of those companies.  Everyone wanted to get in on a 
Microsoft at the ground level.


>>His statement that GCC is "inferior" is also debatable.  GCC is designed
>>primarily to be portable, and efficiency is a secondary consideration.
>>
>
>oh?  So you run NetBSD other than linux because you think it is superior?
>Portability is an attibute of how something does something.  It is not a ranking.
>

Nope.  The primary goal of Linux is not portability.  The primary goal 
of Linux is that Linus have fun.  The primary goal of GCC is 
portability; and by that criteria, it succeeds wonderfully.  I use GCC 
because I know my source will be able to compile on many different 
platforms.

>>GCC has better standards compliance for C; 
>>
>
>I've seen lots of code that GCC eats that it shouldn't.  Most of the problems
>reported for gcc 3.x is that it won't compile things that it did before - GCC
>is not a poster child for standardization.
>

I know of no compiler out there that supports the standards very well. 
 So I suspect you are right; it depends on expectations.

>>it's C++ support is lags a little bit, but is quickly catching up,
>>
>
>yeah... it's only been around for what?  15 years?
>

Yep.  But C++ sucks to start with.  I've not seen too many people 
interested in the C++ end of the compiler.  And that is a drawback of 
free software, both copyleft and BSD/X11; people work in their areas of 
interest.


>>especially with regards to templates.  So, for the design goals of GCC,
>>it is far beyond *any* other compiler.
>>
>
>personally, I think the unix philosophy of small(er) components ends up with
>better results than one monolithic program.
>
Agreed, 100% and wholeheartedly.  I think the GCC approach is flawed in 
that respect.  (And many others.  I don't claim it's a great compiler. 
 I only claim it has achieved the goal of portability.)

>>In fact, I'd suggest that GCC even *created* some compiler markets.  For
>>instance, several years ago, every Unix came with a compiler.  Now,
>>hardly any commercial Unix comes with a compiler; it has been unbundled,
>>and is sold separately.  Sun now makes money off its compiler, when
>>before it was bundled with the OS.  Ask Sun why, and they will say, "Use
>>GCC if you need a free compiler.  Buy ours if you need speed and
>>efficiency."  At least, that was the gist of it when I talked to a Sun
>>engineer.
>>
>
>This is *precisely* the case why the GPL sucks.  So, instead of 10 people paying
>$200 for a compiler (the bundled cc on SunOS sucked.  It was K&R only.  It was there
>because you had to relink the kernel), the gcc makes it so 9 / 10 who can suffer with
>the mediocre compiler does, and the 1 person left now has to pay $2000 for the
>compiler
>(and as a person who DOES buy the compiler, the price has gone up significantly,
>while
>they OS prices has been going down, so it's hard to argue that it doesn't have any
>effect).
>Had it been BSD based, I think things would be different.  Under this scenario, there
>would be incentive to use the BSD base, and make extensions off of it.  So Sun/IBM...
>would still have (presumably) their own version, but the software would be cheaper
>to produce (thus allowing the price to be lower), and the compatibility between the
>compilers would be better.
> 
>
I will not argue with you, as I think the economic analysis (by pretty 
much everyone here, especially me) tends to be amateurish, at best.  But 
I suspect you may be right, in that a BSD-based compiler could reduce 
the amount of work required to produce a new compiler for a new 
platform.  But then again, there's a "not-invented-here" attitude that 
goes with every new architecture; so perhaps not.

>>Some people really like the chaos around the Linux kernel development.
>> I know I prefer the unfettered nature.  And if I wanted a
>>technically-superior kernel, I'd switch to the Hurd, which is a better
>>designed kernel than Linux or any of the BSDs.  And it came right from
>>the bossom of Stallman himself, so it'd piss Brett off even more.
>>
>
>I should point out that proverb: 'what looks good on paper doesn't necessarily
>translate to the real world'.... (or something like that).  Look at NextStep.
>
>-r
>

I'm not sure what you're trying to prove: the NeXT was a great machine, 
years ahead of its time.  NeXTStep was easy to program for; and the NeXT 
interface and OS was well-thought-out.  (It had its problems, to be 
sure.)  It did more than look good on paper; it just looked good, period.

I think the NeXT suffered from the same syndrome BSD suffers from today; 
it's technically superior, but it doesn't get the press.

                - Tony