Time-based Samba4 releases?

James Peach jpeach at samba.org
Wed Feb 1 23:24:21 GMT 2006


On Wed, 2006-02-01 at 11:38 +1100, Andrew Bartlett wrote:
> I've been pondering the prospect of 'time based' Samba4 releases.  In
> particular, the idea of getting TP, alpha and beta releases out on a
> time rather than a feature schedule, so as to keep up interest.
[snip]
> Even so, I think the tech preview shows that we have code worthy of
> broader use, and I think we should aim for a date, with the feeling
> being to defer work rather than push back the release indefinitely.
> (I'm not sure what that date should be, given our wildly varying
> resources however...)

So I thought I'd share my experiences of working in the IRIX release
process in support of this. IRIX releases every three months without
fail.

This is a strict time-based release. If a feature is not ready by the
time a release comes round then it has to slip to the next release. The
tree is continually tested (just like make test). If a bug is found too
close to release time, the offending code is pulled and the release goes
ahead without it.

This has a number of advantages for everyone. From a developer point of
view, I don't have the pressure of having to get feature A done by a
particular deadline, since I know there will be another release window
rolling by real soon. I also have a clear idea of when I should be
pushing features into mainline. Clearly I want to do the disruptive
changes in the first month, and only fix minor bugs in the last month.

For admins and users, the predictable release cycle makes it easier to
plan upgrades, and distro packagers and vendors are also able to plan
their release strategies more easily.

The 3-month release period is useful because it's enough time for major
features to be integrated, but not too long for people to wait for their
favourite bugs to be fixed. I think that 6 weeks is a little too short,
possibly overwhelming testers and users. 6 months is too long - too many
changes will build up in the tree, making it difficult to stabilise.

This is a release strategy that has worked really well for us. To
implement it in Samba might require some changes to how we integrate
code into mainline, but IMHO it would be worth it.

-- 
James Peach | jpeach at samba.org



More information about the samba-technical mailing list