Re: A reply to "Diagnosing SVN"

(Aside: thanks to Jani Averbach for the small correction.)

On Thu, 2004-01-29 at 15:42, Edmund Horner wrote:
> So when are we getting our SQL-backed filesystem? ;-)

I don't know; my personal interest is in developing an FS back end
backed only by the native filesystem, which could be used over remote
filesystems. I have some interesting ideas on this front, but I don't
know how well they will pan out performance-wise.

> Got any recommendations where to start?

I'd start by reading libsvn_fs/structure. I'm guessing that if you're
implementing the FS on top of SQL, you'd probably want to use the same
structure we have there, only instead of "skels" you'd use multiple
columns. Most of the BDB-specific logic is stashed away in the "bdb"
directory and in utils/fs_skels.c.

One of the big missing pieces for this work is a vtable which allows
multiple different FS implementations. Worse, we don't know precisely
where this vtable should live. An SQL implementation might be able to
share a lot more common code with the current BDB implementation than my
native-filesystem implementation can.

---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Thu Jan 29 17:15:40 2004