Re: http://gcc.gnu.org/mirrors.html

[Matthew Langford <langfml at Eng dot Auburn dot EDU>]

On Fri, 14 Dec 2001, Gerald Pfeifer wrote:

> On Fri, 7 Dec 2001, Matthew Langford wrote:
> > And yet, as I mentioned, there must be a gazillion GNU mirror sites.
> > You _know_ they are mirroring, and they are mirroring a gzipped release
> > of your software.
> 
> I assume you missed the fact that
>   http://gcc.gnu.org/gcc-2.95/gcc-2.95.html
>   http://gcc.gnu.org/gcc-2.95/gcc-2.95.1.html
>   http://gcc.gnu.org/gcc-2.95/gcc-2.95.2.html
>   http://gcc.gnu.org/gcc-2.95/gcc-2.95.3.html
> and
>   http://gcc.gnu.org/gcc-3.0/ has been
> have been referring to the GNU mirror sites for years, right?

I see that, and that's exactly what I was requesting, but for the Mirrors
page. Yes, I did miss that.  Problem is, I was looking at the libstdc++
page, which does not have that link.  Nor will you find it if you jump
straight to "Mirrors" to begin downloading (eg, you already know what the
latest release is, don't want to read about what it is/features it
has/bugs it fixes/all about free/open software, just want to get it to
start trying to build).  Nor, if you go to the Installation link, then to
Downloading GCC... "yeah, that's what I want!  Wha...?  It doesn't have
actual downloading information, it just tells me what I get if I were to
download at some future point.  Then I must go to the Releases page..."
and right away I see "mirror sites" and I've once again missed the link to
available North American downloads.

My thought is that for the (most?) common case of wanting to get the
latest gcc source or binaries, minimizing the number of links clicked and
reading required (ie, on the sidebar or at the top of the page) would be
nice.


> > Why not mention _them_ on your downloads page?
> 
> That's a good suggestion, and I will try to add this tomorrow. (Note,
> though, that the GNU site only carries a subset of what's on gcc.gnu.org
> and most of its mirrors, namely current GCC releases.)

Thank you.  Yes, I realize the subset bit, and would think it even proper
if you mentioned this limitation near the link.


> > [...] the poor saps who walked to your ivory tower and volunteered their
> > mirrors?
> 
> Please tone down. Phil is a volunteer, I am a volunteer, and most others
> here are volunteers as well; no money, little honor, few thanks.

I understand, and apologize, in that I probably misconstrued your
response.  If someone had pointed out the links mentioned above, I would
have been somewhat satisfied.  Instead, someone defended the fact that
downloaders were limited to a very lacking mirror list (as in lacking
enough mirrors), which is clearly sub-optimal.  And then defended against
adding to the too-small list, because it was such a privilege to have
unrestricted FTP access.  The overwhelming sense was that the gcc team
could care less if people could reach their software.  Not good.

It didn't help that after several days of trying to compile KDE for
Solaris, I realized that it used too many g++-isms and would not compile
(for me, at least) under the Sun CC compiler--some "file level access"
complaint about a struct with member functions used to instantiate a
subclassed templated class.  Yech.  It's some comfort that gcc 3.x
probably croaks at this, as well as Sun's compiler.  Then I find our gcc
had been compiled without shared libs, so I was going to have to recompile
gcc, as well as recompile a ton of other libs and add-ons (libxml,
libxslt, Qt, then kdexxx).  I wanted to get gcc 2.95.3 as quickly
as possible to get on with it.

> > I'll ignore the rest of your flames and just explain this:  gcc.gnu.org
> > is extremely overloaded.  For that reason anonymous FTP has been severely
> > restricted (and I think it's been disallowed completely on some of the
> > more popular projects).  We're trying to move anonymous CVS access off
> > onto another system also.
> 
> Yup, and that seems to work fine now; I already have a patch and will
> update the corresponding GCC page tomorrow. (Again, Matthew, also this
> has been handled by volunteers exclusively.)

Understood.  Please understand me:  for your efforts to have meaning, your
software needs to be distributed to the people who want to use it.  It is
in your best interest to work around the issues surrounding gcc.gnu.org
and the limits on gcc-specific mirrors, and provide as easy a path as
possible for obtaining your software.  That was why I took the time to
send you a note about the problem.


--
MattLangford