This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: CVS using a proxy ?
- To: kingdon at cyclic dot com
- Subject: Re: CVS using a proxy ?
- From: Hans-Peter Nilsson <hans-peter dot nilsson at axis dot com>
- Date: Sat, 18 Jul 1998 22:21:44 +0200 (MET DST)
- CC: richard dot earnshaw at arm dot com, hans-peter dot nilsson at axis dot com, nicolas at bnp-eng dot remcomp dot com, egcs at cygnus dot com, larry dot jones at sdrc dot com
> Date: Sat, 18 Jul 1998 14:57:01 -0400
> From: Jim Kingdon <kingdon@cyclic.com>
> References: <199807181807.TAA24493@sun52.NIS.cambridge>
> Richard Earnshaw <rearnsha@arm.com> writes:
> > But there is no need to use a special (socksified) variant of cvs.
> > For machines that support dynamic linking there is a script that comes
> > with socks that allows you to change the inet protocol stack used at
> > run time.
Yes, I vitally forgot to mention "runsocks"; that also solves the problem that
you don't *have* to compile cvs yourself, etc. A lot simpler and less
intrusive on the behalf of cvs, when it works.
> This is partly just because I am a lazy application writer and don't
> want to have to update my application for the 3 zillion different
> authentication/connection schemes out there. But it is also The Right
> Thing(TM) for network security -- you don't want the security of your
> network to be at the mercy of whether CVS (and every other
> application) has gotten things right.
Laziness has many faces; my laziness expressed itself in that I checked that
linking with socks would DTRT, so I don't have to remember and type
"runsocks" before "cvs" when running it manually ;-)
And of course, you know that you call it "cvs", *wherever the current
repository is*. (That, and runsocks wasn't correctly installed/didn't DTRT
where I first tried it, so I got a bad first impression. 1/2 :-)
> I've updated the SOCKS information at
> http://www.cyclic.com/cvs/dev-net.html to suggest runsocks.
I agree that this solution is good (a lot better than suggesting a source
patch), if you see the firewall connectivity solution as an add-on, or for
"personal" use; for site cvs installations by a sysadmin I believe a
socksified cvs is better. Hmm, on second thought, they'd probably hide cvs
and install a script in its place that said "runsocks cvs.hidden $*"...
And now you probably don't want a patch to cvs for --with-socks support?...
(I'll stop CC:ing egcs after this, sorry for the diversion).
brgds, H-P