Re: SoC: Path-based authz for Svnserve

Garrett Rooney wrote:
> I'd rather you just stat the file to see if it's been modified or
> something like that, requiring the user to send a signal to the process
> seems rather error prone.

It is error prone. I'll go with what Greg suggests: reread the file on
each new client connection.

> Some way to indicate that a part of the tree can be accessed by a user
> that is not logged in. I don't think mod_authz_svn can do that, but
> svnserve might want to.

Is this more badly needed than wildcards on paths? It seems those two
enhancements are mutually exclusive, or at least having both involves
some complexities.

> I'd say go without it for the first pass, just reading it in at the
> start of a session, then after you've got it working look at how much of
> a speed improvement there is when caching is added.

As Greg said, there is the risk of any caching I add ending up being
slower than the actual parsing. It'll all depend on how fast the parsing
into hierarchical hashes makes the subsequent analysis. Anyhow, I'll
look this up after all the rest is done, so there is time.

If all goes well, the first patch moving the authz routines into
libsvn_repos and updating mod_authz_svn accordingly should be posted
within a few days (but of course all will not go well ;).

- Dave.

---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Thu Jun 30 20:01:40 2005