[svn.haxx.se] · SVN Dev · SVN Users · SVN Org · TSVN Dev · TSVN Users · Subclipse Dev · Subclipse Users · this month's index

Re: FSFS as default for svn 1.2?

From: John Szakmeister <john_at_szakmeister.net>
Date: 2005-03-22 12:06:07 CET

On Tuesday 22 March 2005 05:32, Erik Huelsmann wrote:
> > Ben Collins-Sussman wrote:
> > > My only caveat: if we do decide to switch to FSFS as the default,
> > > we need to be very careful about how we announce/describe that
> > > fact. It might be very easy for people to think that we're
> > > dropping BDB support (or deprecating it, phasing it out) when in
> > > fact the exact opposite is true: Sleepycat is working to make BDB
> > > better, and we're working to make better use of it. It's just that
> > > none of that will be visible yet in 1.2. :-/
> >
> > Reading:
> >
> > http://svnbook.red-bean.com/en/1.1/ch05.html#svn-ch-5-sect-1.3
> >
> > there don't actually seem to be any really compelling reasons to
> > use BDB. If FSFS becomes the default, in what circumstances would
> > someone want to convert their repository back to BDB?
> >
> > I guess playing the devils advocate here, if FSFS is mature enough
> > now, and BDB offers few advantages and has several disadvantages,
> > why continue to support BDB?
> Because we'd have to release 2.0 to be able to drop it...
> But also: there are some operations which are faster in the bdb setup.
> In the 1.x series 'svn blame' is a lot slower for FSFS than it is for
> bdb. (This has been addressed for 1.2 though.)
> There are some others too. Although it seems to scale better than we
> originally thought it would / the use case it was designed for.

* Ease of creating indexes

We don't know right now what we might add in the future. It could be that
those things don't lend itself to the FSFS implementation very well, and
is more easily computed by the BDB backend.

* Maturity

It's been out since the beginning, and while it has it's share of
constraints when running in a multiple-access environment, it has years
of use underneath it's belt.

* Ensuring that we do our due diligence in trying to have an API that
supports multiple backends (an SQL backend will be a reality... one day)

This is more for the developers that than the users, but still important,

For the foreseeable future, I don't think we should even consider dropping
support for BDB. However, I do think that making FSFS the default will
provide a better up front user experience, and would vote for doing that.


To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Tue Mar 22 12:10:34 2005

This is an archived mail posted to the Subversion Dev mailing list.

This site is subject to the Apache Privacy Policy and the Apache Public Forum Archive Policy.