:: Re: [Dng] rumors on RMS about syste…
トップ ページ
このメッセージを削除
このメッセージに返信
著者: Jude Nelson
日付:  
To: devuan.kn
CC: dng@lists.dyne.org
題目: Re: [Dng] rumors on RMS about systemd at libreplanet
> GCC was deliberately making things interdepend on each other, even
> without technical reasons, simply to prevent commercial entities to
> replace the e.g. front-end of the compiler with some proprietary code
> and then have that use the GPL backend. This would enable a new,
> proprietary language to leverage all the optimizations gcc has. So
> this prevents what Apple does with swift on llvm right now, and I
> understand the reasoning behind that decision, even though I regret it
> since it prevents us from having many valuable tools for code analysis
> and refactoring.


GCC was also built at a time when proprietary compilers were the norm, and
notions of "free software" and "open source" did not exist outside a few
close-nit circles. What RMS did with GCC's design was make it very
difficult to create and distribute proprietary extensions to GCC without
violating the GPL. He did so by designing GCC such that the only supported
way to add new extensions was to *statically* link them into GCC at
compile-time. For example, NeXT tried to distribute a proprietary
Objective-C module as a .o file that had to be linked into GCC by the
developer, but were ultimately compelled by the GPL to release the code.

It's worth pointing out that GCC has since become a lot less monolithic (it
was still modular internally, however), in part due to competition with
LLVM/Clang, in part because it make interoperability easier, and in part
because the market expects open source compilers.

Not sure how this is related to systemd exactly, but suggesting that we
prioritize LLVM/Clang over GCC because we prioritized sysvinit and other
daemons over systemd would be disingenuous. Systemd has the opposite
trajectory, for one--it's becoming *more* monolithic, not less. Also, one
can have LLVM/Clang and GCC installed and running at the same time without
loss of functionality. This is not true for systemd versus any other init
system, since there can be only one PID 1.

-Jude

On Mon, Mar 23, 2015 at 7:17 AM, <devuan.kn@???> wrote:

> On Mon, Mar 23, 2015 at 11:18 AM, Jaromil - jaromil@???
> <devuan.kn.0edf9dfcba.jaromil#dyne.org@???> wrote:
> > On Mon, 23 Mar 2015, KatolaZ wrote:
> >> This sounds strange and new at the same time, since GCC was indeed
> >> designed to be portable and ported to several architectures since from
> >> the beginning. Do you have any quote by RMS or by any guy who has
> >> contributed to gcc to support your statement? I don't think that GCC
> >> is a hairball, to be honest.
>
> Oh, a hairball can be portable in its entirety:-) It is just hard to
> use parts of it.
>
> GCC was deliberately making things interdepend on each other, even
> without technical reasons, simply to prevent commercial entities to
> replace the e.g. front-end of the compiler with some proprietary code
> and then have that use the GPL backend. This would enable a new,
> proprietary language to leverage all the optimizations gcc has. So
> this prevents what Apple does with swift on llvm right now, and I
> understand the reasoning behind that decision, even though I regret it
> since it prevents us from having many valuable tools for code analysis
> and refactoring.
>
> https://lwn.net/Articles/629259/ covers the most recent flare-up when
> somebody wanted to make the AST of GCC accessible.
>
> > This is not the first post in which this spamgourmet account is
> > spreading FUD. We may need to react to this beyond argumenting.
>
> I don't see anything I said on this list to be related to be spreading
> Fear, Uncertainty or Doubt.
>
> I did argue for some of things I think are sensible, just like
> everybody else here. All of them can be implemented _without_ systemd
> (and I have or had them running that way). Yes, one of the ideas I
> like was first proposed by Lennart, but is this about building a new
> distribution or about holding a grudge?
>
> If it is the later, then you won't need to bother to ban me.
>
> _______________________________________________
> Dng mailing list
> Dng@???
> https://mailinglists.dyne.org/cgi-bin/mailman/listinfo/dng
>