This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: anonymous CVS stuff
- To: egcs at egcs dot cygnus dot com
- Subject: Re: anonymous CVS stuff
- From: Marc Espie <espie at quatramaran dot ens dot fr>
- Date: Mon, 26 Jul 1999 13:41:19 +0200
- Organization: Ecole Normale Superieure de Paris
- References: <3181.932938192@upchuck.cygnus.com>
In article <m3hfmsxdid.fsf@geoffk.wattle.id.au> you write:
>Would it be useful if the FTP server had a .tar.bz file of a daily
>checked-out CVS tree (with CVS/ directories and everything)? That
>way, people could avoid using CVS for initial checkouts of the
>mainline, which should reduce the load (it would at least eliminate
>the cost of compressing all those files).
>The idea would be that people could just do
><ftp "egcs-cvs-daily.tar.bz">
>bunzip2 < egcs-cvs-daily.tar.bz | tar xf -
>cd egcs
>cvs login
>cvs -z9 update
Not really that nice in terms of bandwidth, but it gives me an idea.
What about rsync ?
If you set up an rsync anonymous server, and provide a checked out CVS
tree, it may be useful for people who mostly want the current source
and a way to do diffs afterwards.
There are two major ways to do that: 1/ either you provide a full CVS tree (but
then, it can be slow, as rsync still has to walk that tree) OR 2/ you provide
a simple tar archive, NOT COMPRESSED, as rsync will compress it by itself.
1/ is highly convenient for end users, as they can even update part of their
tree.
2/ yields a much lower load on the server, and needs that people
be able to cache a tar archive somewhere.
In any case, 1/ still is much faster and more system-friendly than CVS, but it
can't merge local and distant changes... (yep, there ought to be a gotcha).
Rsync plug: rsync.samba.org, from the same guys who wrote samba.
rsync knows how to synchronize large files over networks, by mostly
transferring the parts that changed, and a few checksums.