On Feb 17, 2009, at 8:18 PM, Talden wrote:
>> There are a number of new features in 1.6, some of which are user-
>> visible, others which are not as visible. The Subversion community
>> trying hard to do more frequent minor releases, in an effort to cut
>> down on the amount of new stuff in each release. This makes
>> development more manageable, decreases the new testing surface area,
>> and hopefully gives users access to more features more frequently.
>> Compared to the almost 2-year cycle for 1.5, 9 months for 1.6 is
> Hyrum, this is the bit that gets me. The WC-NG effort is fantastic
> but is real, true effort with a lot of testing. If the 1.6 cycle is
> considered fast and 1.6 is not yet final then we're talking 9 months
> (I saw 6 mentioned, that sounds hopeful) to get this improvement.
Actually, 1.5 was an aberration from the typical release process,
normal release cycles have historically been between 6 and 9 months.
We're hopefully of getting 1.7 released by the end of the summer. WC-
NG development is fortunate in that we can use the existing blackbox
testsuite to vet bugs as they crop up, instead of relying upon ad hoc
testing further in the development cycle.
> Can we not have WC-NG in 1.8 and get a 1.7 ASAP after 1.6 that has a
> fix for this locking performance issue.
That was briefly discussed among a couple of the developers earlier
today, but the WC-NG work is now on trunk, so it will be in the next
release. Early testing and bugfixing is something you can do that
will help get that release here sooner.
It should also be note again that the purported patch is not complete,
hasn't been implemented or tested, and any performance gains have not
been thoroughly quantified. If development experience with the
current working copy library is any indicator, such "trivial" patches
tend to be more work than they appear at first blush.
> It seems we're either holding off just because the release number
> (1.7) has already been assigned for WC-NG or because we're holding
> back this fix to make the 1.6 -> WC-NG performance comparison look
The release number may be part of it, but not a large part. I can
assure you that there's no cabal in a smoke-filled back room
attempting to game performance measurements from one version of
Subversion to another.
> On my laptop, defragged and as disk-cached as possible, an empty
> update from a LAN server takes me 1m30s with svn://... Doing an pull
> into a branch-with-working-tree in Bazaar takes 30s - and Bazaar is
> considered one of the slower DVCS tools. In nine months there might
> be a few users to show that nice and shiny WC-NG to.
That may be true, but as has been discussed ad nauseum in other
forums, Subversion and DVCS tools fill somewhat different niches in
the same ecosystem. Sure there are "competing" customers, but a large
number of Subversion users will be around in several months...and
those are largely the ones driving WC-NG development.
> Is there really no way this number of voices here can't shift the
> project plan and get an interim fix in - redundant once WC-NG arrives
> (which I am excited about, don't you worry).
The best way to get voices heard is with code. That's the way an open
source project works: the people who contribute are the ones who set
the direction the project will take. It's not that the developers are
ignoring the user community (this thread is good evidence of that)
it's just that we have our own agendas, personal or corporate, and
those take precedence.
Again the invitation: join the fun, write some code, and help wc-ng
get here just a little bit sooner!
PS - I think I've said all I can on this subject, so I'm going to sit
back and enjoy the view.
To unsubscribe from this discussion, e-mail: [users-unsubscribe_at_subversion.tigris.org].
Received on 2009-02-18 03:42:41 CET