Bryan,
This proposal to change to the current milestone/release numbering
scheme was initiated by me in the first place. And I had a number of
reasons for that as listed below by Heikki. The current numbering scheme
suffers from drawbacks which mostly affect release management and QA.
See my comments inline..
Bryan Stearns wrote:
-1 to this 'forward' numbering scheme: I voiced an opinion earlier
today in IRC that this numbering scheme is confusing, but Heikki said
it had already been decided.
Every project I've worked on works the old way: intermediate builds
have a number based on the previous major release, with something
incremental added to it. The only exceptions to this are
release-candidate builds (which aren't done unless they're _really_
release candidates: the only way a build gets tagged with
1.0-RC-something is if people believe it might actually be the build
that gets tagged 1.0).
It seems wrong that 0.7 is going to sort before all of the
intermediate builds that lead up to it.
(more below)
...Bryan
Heikki Toivonen wrote:
Many people have complained that the current milestone numbering scheme
is confusing and unworkable. The current system is simply ascending
numbers with major, minor and micro revisions, where the micro revision
is padded with a leading zero if it is less than 10 (to help sorting).
So for example, these milestones would happen in this order: 0.5,
0.5.01, 0.5.02, 0.6.
Some problems that have been mentioned with this system (I may be
forgetting some):
* Doing a bug fix release of a milestone/release would need to add
fourth group of numbers, which seems excessive. For example, if we'd
need a bug fix of 0.6 release, it would have to be 0.6.0.1 to
distinguish from 0.6.01 milestone on the trunk (and even then there
might be confusion because of the leading zero).
I don't see how the confusing forward-numbering scheme affects this.
How does it solve this problem?
[Aparna]: The forward numbering scheme does indeed solve this problem
because the trunk would be numbered 0.7 onwards whereas the branch
releases will continue as 0.6.01, 0.6.02 etc. This would be lot less
confusing than having trunk on versions like 0.6.01 and the branch on
0.6.0.1
* It's hard to talk about a milestone with these major, minor and micro
numbers. The informal way to talk about the milestones has been to say
m<some number>, which means the the <some number> micro revision in the
current release cycle. So currently m8 would mean 0.5.08. But this is
informal, and changes meaning once we switch focus to the next release.
Again, the forward-numbering idea still has us talking using major,
minor, and micro numbers. How does it solve this problem?
[Aparna]: Well atleast we will all be on the same page when saying we
are in m8 milestone. Currently we say we are doing milestone m8 but the
builds are called 0.5.08. We actually had problems where people couldn't
connect the two and led to some miscommunication. This is mostly to keep
terminology consistent in our communication.
* Some people would like to use the Bugzilla version field to mark the
version of Chandler in which the bug was found. This is clear enough
with a release, but unclear with the current numbering when working
towards a release. Should the version currently be 0.5 (the previous
release) or 0.6 (the release we are working towards)?
The bugzilla "build identifier" field should be used to record the
specific build where the bug was found. The version number field is
too limited to be useful, and the guided bug-entry page doesn't even
ask for it separately anyway... further, I again don't see how
numbering forward helps in this case: reporters need to record the
specific build in which they found the bug no matter what nomenclature
we use.
[Aparna]: This is one of the main driving forces for this proposal.
Having the build identifier is useful in some sense but it doesn't in
any way tell you which release version this bug was found in. For e.g.
our current build identifiers look like :
Chandler_win_20051129191940.exe but it would also be helpful to know
which release we were working on when this bug was found. This
information would be very useful for QA.
Since Version field is one of the fields for in bug entry form most
people(internal and external) are likely to set the field to some value
and currently they are either being set to 0.5 (since that's what our
build numbers say for milestones) or 0.6 because that's the release
version we are working on. So there's tremendous confusion there. This
would be even more critical once we branch we have bugs being filed on
the 0.6 branch v/s the 0.7 trunk. Also, setting the version field
correctly helps us to track number of bugs filed/resolved/verified for a
specific release and do some data analysis based on that. Currently we
have to resort to querying based on dates.
Anyways, adding the version number is just more information in the bug
report, which if not to anyone else, is helpful to QA.
I am definitely voting +1 on this.
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _
Open Source Applications Foundation "Dev" mailing list
http://lists.osafoundation.org/mailman/listinfo/dev
