#opendaylight-meeting: tws
Meeting started by phrobb at 18:00:54 UTC
(full logs).
Meeting summary
- Today's meeting is on Boron Planning (phrobb, 18:02:40)
- colindixon says at this point we need to decide
(soon) whether we want to do a more normal Boron release and look at
fast-phased in Carbon, or try to do it in Boron (colindixon,
18:08:08)
- skitt says that his take is that should be
mostly up to offset 0 proejcts (colindixon,
18:08:59)
- colindixon notes that we don't really have any
of them (colindixon,
18:10:00)
- vishnoianil asks if we have a good writeup of
pros and cons of the two approaches (colindixon,
18:10:16)
- colindixon says he doesn't think they're
writtne down concretely except in minutes (colindixon,
18:12:19)
- Anil asks if there is a detailed write up of
rational for the change and new issues expected as a result of such
a change. colindixon notes that there isn't really document with
this information... the general rational for the change is a move to
agile, faster releases so that projects have less time to get into
trouble related to integration/dependency issues. Also, the method
is trying to more loosely couple (phrobb,
18:13:53)
- colindixon says he thinks there are two key
independent pros: (1) shorter releases are easier to manage in
general, (2) corss-offset dependencies would be on released versions
helping to avoid transient breakages (colindixon,
18:14:03)
- edwarnicke notes that there's a thrid:
integrating with external projects (in both directlions) has to wait
less time if you can release more often, e.g., don't have to wait up
to 6 months to integrate (colindixon,
18:14:54)
- colindixon says the biggest cons seem to be (1)
a potential explosion of branches, (2) you have to wait until
upstream projects release to get bug fixes, (2) unknown
unknowns (colindixon,
18:17:34)
- edwarnicke says that one way you can avoid an
explosion of branches is to not have all releases be
long-term-support, colindixon notes that we talked about this last
time (colindixon,
18:18:28)
- edwarnicke asks if he could prepare a
parameterzied proposal, many people say sure but would like to see a
spreadsheet they can tweak and at least one concrete example
(colindixon,
18:19:19)
- skitt says that a lot of this complexity comes
down how long releases are supported to external customers and how
fast downstream projects are willing to move to newer
releases (colindixon,
18:20:41)
- skitt also notes that semantic versioning might
help to make sure projects were willing to do some upgrades (e.g.,
safe ones) more often (colindixon,
18:23:03)
- vishnoianil asks the obvious question of
whether there is too much overhead in a 2 month release (colindixon,
18:25:10)
- edwarnicke says yes, it's possible, other
things get easier because you're coordinating with fewer other
projects (colindixon,
18:25:30)
- another key con you potentially have to deal
with is version skew (colindixon,
18:28:43)
- ACTION: edwarnicke to
write up a list of pros/cons and maybe a parameterized list of this
plan (colindixon,
18:29:25)
- skitt talks about how you might manage overhead
by not delivering new functionaliy every release, e.g., only
shipping new functionality only when it's ready, you can do this
with topic branches or unpublished karaf features or other
ways (colindixon,
18:30:34)
- vishnoianil notes that there is risk with
putting features that might not be completed, then you're sort of
stuck with half done things (colindixon,
18:32:37)
- edwarnicke points out (and colindixon and
vishnoianil mostly agree) that topic branches have their own
issues (colindixon,
18:33:07)
- skitt notes that to handle version skew we need
new tooling, and then likely a carrot and stick approach to getting
projects to upgrade to new versions (colindixon,
18:35:48)
- skitt notes that might be new functionality
(carrot) and lack of support (stick) and maybe mandatory version
(ranges) to be part of an "OpenDaylight" release vs. a project
release (colindixon,
18:36:44)
- skitt and colindixon talk about the
complexities of trying to deal with version ranges (colindixon,
18:41:14)
- skitt notes that there's source-code level
compatibility that won't necessarily guarantee binary compatibility
for things like yantools version they were released with
(colindixon,
18:41:55)
- LuisGomez says that depending on multiple
different releases makes testing harder (which it does) and mabye we
should just make everyone use the latest LTS when we release
it (colindixon,
18:44:05)
- skitt notes that we need to define what it
means to be part of the ODL community and his hope would be that if
something critical is in need of help, then they would get the help
they asked for (colindixon,
18:45:10)
- anipbu asks if projects can choose to skip a
simultaneous release. colindixon says as our governance is written
projects are allowed to participate or not as they see fit
(colindixon,
18:51:50)
- colindixon says he really hopes that we make
participating in a release simple enough that people don't worry
about choosing to opt-in (colindixon,
18:52:26)
- colindixon notes that his experience is that
the time somebody says "I won't participate in this one, I'll do the
next one" is often the last time they make progress becaue it gets
harder to integrate (colindixon,
18:54:42)
- anipbu asks about how SRs would work, would we
still have them (colindixon,
18:55:57)
- colindixon says he thinks we'd need to so that
we could fix bugs for downstream projects in betwen releases, and
they might be more often beaause there won't be as much direct
integration testing without snapshot integration (colindixon,
18:58:24)
- skitt noted way up above that upgrading
versions is really complicated here because we have at least 4
versions: yang revisions, osgi versions, maven versions, and karaf
feature versions (colindixon,
18:59:18)
- next steps (colindixon, 19:01:31)
- the next steps seem to be (1) writing this up
and (2) figuring out if offset 0/kernel projects want to do it for
boron (colindixon,
19:01:49)
Meeting ended at 19:01:54 UTC
(full logs).
Action items
- edwarnicke to write up a list of pros/cons and maybe a parameterized list of this plan
Action items, by person
- edwarnicke
- edwarnicke to write up a list of pros/cons and maybe a parameterized list of this plan
People present (lines said)
- colindixon (39)
- phrobb (5)
- odl_meetbot (5)
- edwarnicke (4)
- zxiiro (3)
- anipbu (0)
- CaseyODL (0)
Generated by MeetBot 0.1.4.