rsync -> cvs down?
On Mon, 2004-07-19 at 14:57, Martin Maechler wrote:
"Marc" == Marc Schwartz <MSchwartz@medanalytics.com>
on Mon, 19 Jul 2004 13:22:43 -0500 writes:
Marc> Uwe,
Marc> That did it. Using https: I am now able to do a checkout.
Marc> It seems to be slow at the moment, but the files are coming through.
---------------------------------
that should have improved now.
The default apache2 configuration (for RH Enterprise) had
'KeepAlive Off' which I now have replaced with 'On'.
For this to take effect, I had to restart the server --
this pretty brutally terminates all running svn requests, which
in this case lead to the need for "svnadmin recover"ing the
archive... Well, well, we're getting there eventually.
If anybody has good experiences to share about
Apache performance tweaking, please let me hear.
But note that the idea (of the server setup) really was to serve
R-core (and maybe ESS-core and maybe some really small few-person
collaboration projects). If too many people (such as "hundreds
of R-devel readers") are going at the server it will become
pretty unusable, and I will have to make it accessible only
"non-anonymously" {and find another one doing rsync...} eventually.
Martin Maechler
Martin, To play 'devil's advocate', if you want to limit access for the group(s) that you have identified above, you may wish to consider having references to the use of rsync/CVS (now subversion) access removed from the CRAN Source Code Page (ie. http://cran.r-project.org/sources.html) and from pages 1 - 2 of the R-admin manual. The rest of us can always download the daily stable, patched or devel tarballs periodically, though of course losing the 'incremental download' functionality by doing so. I am not sure that most of us folks need to have access to hourly updated source code. It's a question of being proactive or reactive... Of course, now that I have stepped face first into it today, it's in the archives... ;-) Best regards, Marc