Re: GCC is 25 years old today
Thu, 29 Mar 2012 12:30:09 -0000

          From comp.compilers

Related articles
[3 earlier articles]
Re: GCC is 25 years old today (2012-03-27)
Re: GCC is 25 years old today (glen herrmannsfeldt) (2012-03-28)
Re: GCC is 25 years old today (Dmitry A. Kazakov) (2012-03-28)
Re: GCC is 25 years old today (Rui Maciel) (2012-03-28)
Re: GCC is 25 years old today (BGB) (2012-03-28)
Re: GCC is 25 years old today (Hans-Peter Diettrich) (2012-03-29)
Re: GCC is 25 years old today (2012-03-29)
Re: GCC is 25 years old today (Joshua Cranmer) (2012-03-29)
Re: GCC is 25 years old today (BGB) (2012-03-29)
Re: GCC is 25 years old today (Hans-Peter Diettrich) (2012-03-30)
Re: GCC is 25 years old today (Dmitry A. Kazakov) (2012-03-30)
Re: GCC is 25 years old today (BGB) (2012-03-30)
Re: GCC is 25 years old today (BGB) (2012-03-30)
[5 later articles]
| List of all articles for this month |

Newsgroups: comp.compilers
Date: Thu, 29 Mar 2012 12:30:09 -0000
Organization: Compilers Central
References: 12-03-061
Keywords: GCC, Ada, history
Posted-Date: 30 Mar 2012 02:40:02 EDT

Dmitry wrote:

> compilers wrote
> > As you know Adacore still makes plenty of money.
> First GNAT is GCC. Second AdaCore does not sell it. GNAT is free. AdaCore
> sells support and an escape from the GPL-jail.

No, I think that is not accurate. Yes, GNAT is GCC, sortof,
eventually, maybe. But to get the "integrity assured" validated
compiler suite you have to pay, last time we heard, 20,000 USD for a
package of (I think) 5 seats. It's disingenous to say you're only
buying support, because Adacore owns the upstream and provides fixes
that GCC eventually adopts and there is no time frame for this. It's
not the same code, and it's not the same tools that GCC provides. At
the end of the day does anyone really think the sales of the "support"
aren't what's funding GNAT? I'm not sure how they get away with dual
licensing if the code is the same but IANAL. I am somebody who thinks
something is rotten in Denmark though. It would be a lot more honest
as far as I am concerned, to just say they're selling it. Escaping
from the GPL-jail is worth something, but you can't call that support.
It's called buying something. Come on, guys!

> > So does Green Hills.
> Yes, compilers for embedded systems are not completely dead. Just because
> new processors come and die quicker than the process of suffocation fastens
> its grip. In fact it is the BSP people are buying, not the compilers.

Is embedded all Green Hills and Aonix and the other Ada vendors do?
They are also offering UNIX and Windows targets and they are not free,
not by a long shot. There are another 2 or 3 Ada vendors that sell for
UNIX and Windows targets I found in a quick web search.

Intel sells Fortran and it is supposed to be excellent, because of
Intel's inside knowledge of its own architecture. The Linux versions
are usually free for academic use, but Windows has no free version and
no academic license. They also sell a C/C++ compiler.

If you look you can find a bunch more companies selling compilers.
Apparently there are still enough people around who don't like GCC even
without the GPL contamination who are willing to pay good money for a compiler.

> And for a given board there is normally only one compiler +/- GCC. This is
> not yet a market. If it became one, it would suffer the same fate, the
> market of PC compilers did. The point is that, as with the case of AdaCore,
> embedded is more about selling services than about selling compilers.

That's the way they position it, because of the GPL dishonesty. But I say if
you can't get the code free without paying for support (or waiting who knows
how long for fixes that may never come) then yes, they're selling it. This
kind of flim flam and hair splitting isn't helpful. You well know several
people have expressed interest on comp.lang.ada in buying a developer
license for a non-GPL GNAT and there was either dead silence or someone
saying Adacore is not interested in small fish, even without support.

> > There are several significant compiler companies. Not as many as in
> > the 1970s and not on as many platforms but there are still some doing a
> > pretty good business.
> How could it become less in the era of mobile computing, GPUs, concurrent,
> networking systems? When computers are infesting everything except for
> maybe door mats (yet to come)?

Because in the early days people paid for everything. It was only after UNIX
with a healthy dose of Stallman convinced the masses of the values of
socialism and putting their hands in everyone's pockets but not letting
anyone's hands in their pockets (funny how that works) people started being
a lot less willing to pay for compilers. Think of all the dozens of pre-PC
computers like the Atari, Commodore, TRS-80, the list goes on and on, there
were dozens of compilers sold for every platform and hundreds of
companies. Borland was built on that landscape.

There are less companies selling compilers now because of what I wrote above
and because the vast majority of the new platforms are using good old C++
(proprietary or otherwise) or Java or JVM languages and they can't sell that
stuff. The runtime and development tools are either given away for nothing to
encourage people to write code for those platforms or packaged just for the
vendor and aren't released at all. GPU stuff is all giveaway to get you to
buy their products. You can't say it's free, although technically it is,
because it's really part of the cost of buying and selling the platform. It
doesn't help you to have nVidia's SDK if it doesn't work on the other guy's
GPU platform. It's only valuable to you if you own nVidia's cards, etc.

[Ahem. In the early days software was all free. It wasn't even
possible to copyright it. Companies started selling it in the late
1960s partly because so realized there was a market for specialized
programs (the first was one that drew the flowcharts that managers
demanded from the code the programmers had already written), and
a legal settlement led IBM to unbundle their system software. But
the ancient free version of Fortran H still generates amazingly
good code. -John]

Post a followup to this message

Return to the comp.compilers page.
Search the comp.compilers archives again.