ibiblio SVN service architecture

Greetings-

Ibiblio is in the process of desining an svn service, and I thought I would
send our current setup to the list to see if any folks had feedback.

We are deploying this service side by side with trac. I'm pretty sure that
the two can live in absolutely perfect harmony, but if it comes down to it,
we will want to use subversion best practices before trac conveniences.
Furthermore, many of the projects hosted might not use trac, but only
subversion.

We are currently providing svn services for 2 internal projects (
http://source.ibiblio.org/svn/lyceum/, http://source.ibiblio.org/svn/osprey).
We anticipate that we might provide hosting for 10-20 additional projects
over the next year. Ideally the system would be able to scale to hundreds.

I've left the trac conf code in the relevant files, I don't think it will be
distracting. Feel free to ignore it, or if you have any comments they are
very welcome.

Here is one of the projects that we already have running:

File structure:

   - The svn repositories are in /srv/svn/repos/. So for example,
   /srv/svn/repos/lyceum. Each repository is what you would expect, the result
   of running `svnadmin create projectname`.
   - The digest password files (also used by trac) are in
   /srv/svn/administrative/. So for example, /srv/svn/administrative/lyceum is
   a file containing HTTP auth Username:Realm:Password entries.

Apache configuration:

   - httpd.conf has LoadModule dav_svn_module modules/mod_dav_svn.so as
   you would expect, and then includes source_ibiblio.conf
   - source_ibiblio.conf has directorives for ALL svn and trac projects.
   It is attached. As you can see, it includes a subdirectory of project config
   files
   - An example of one of these, lyceum.conf, is attached.

Questions:

   1. what do you think of the file structure in general? are there any
   other config files we might need to use in the future that we should design
   a proper place for now?
   2. we haven't considered fine-grained access control at all. i don't
   know anything about this and intend to read about the details of how it
   works, BUT at a more abstract level:
   1. if i remember correctly there is more than one way to do access
      control. is one considered supperior to the other?
      2. what additional config files will be required? where should
      they go?
      3. what apache configuration directives can go in the general
      file, and what will need to go in each project file?

I realize that is a lot, but I thought some of you out there might enjoy
considering the design of such a system. Any feedback is greatly
appreciated!

Cheers,
John

----
aim/yim/msn/jabber.org: johnjosephbachir
713.494.2704
irc://irc.freenode.net/lyceum
http://lyceum.ibiblio.org/
http://blog.johnjosephbachir.org/

---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe@subversion.tigris.org
For additional commands, e-mail: users-help@subversion.tigris.org