#opendaylight-meeting: tws
Meeting started by colindixon at 17:01:48 UTC
(full logs).
Meeting summary
- agenda bashing (colindixon, 17:01:48)
- today we'll cover the OpenFlow projects
organization and future (colindixon,
17:02:10)
- https://wiki.opendaylight.org/index.php?title=Tech_Work_Stream:Main&oldid=46200#Upcoming_Meeting_Agendas
agenda (colindixon,
17:02:29)
- https://lists.opendaylight.org/pipermail/discuss/2016-May/thread.html#6475
thread on this topic (colindixon,
17:02:54)
- https://docs.google.com/presentation/d/1J1ddSWCGXhbz94uNMM8xzO_ssrry3OebV7pgQWKDkrs/edit
the presentation Abhijit will be using (colindixon,
17:03:07)
- openflow strategic items (organization of projects and future) (colindixon, 17:06:21)
- two key items: (a) reroganizing different
openflow projects and (b) OpenFlow 1.5 support (colindixon,
17:06:51)
- apparently, we're focusing on (a) today, but
less so (if at all) on OpenFlow 1.5 (colindixon,
17:07:10)
- we have lots of OpenFlow projects (colindixon,
17:07:40)
- tightly entwined: OpenFlow Protocol Library
(openflowjava), openflowplugin, circuit switching extensions, and
EPC extensions (colindixon,
17:08:14)
- less entwined: TTPs and OF-CONFIG (colindixon,
17:08:23)
- current issues: (colindixon,
17:09:00)
- 1.) some performance cost for double
translation of models (once for openflowplugin and again for
openflowjava) (colindixon,
17:09:04)
- 2.) patches often need to go into both the
openflowplugin and openflowjava project (colindixon,
17:09:25)
- 3.) fewer resources available these days for
OpenFLow-related projects (colindixon,
17:10:00)
- 4.) confusing to people who are looking for the
"openflow support" in OpenDaylight (colindixon,
17:10:28)
- abhijitkumbhare mentions the cost of having two
different models at two different levels: plugin and protoco
library (colindixon,
17:14:11)
- colindixon asks if the performacne impact has
been quantified, abhijitkumbhare says no, not right now (colindixon,
17:14:30)
- abhijitkumbhare talks about extension style
projects, abhijitkumbhare says that ODL forge would be one place to
house all the extension projects (colindixon,
17:18:23)
- vishnoianil asks about maybe using the
top-level project and project managmenent committee (colindixon,
17:18:49)
- colindixon points out that the integration
projects currently do this, but without needing to be top-level or
have a PMC (colindixon,
17:20:21)
- colindixon notes that ODL forge doesn't exit
right now (colindixon,
17:20:44)
- rovarga says the biggest issue is really how
these extensions are maintained and released, e.g., if it's never
released, there's no risk that we have to support it (colindixon,
17:24:06)
- if we release it and do so as part of
openflowpluging, then we have to maintain it (colindixon,
17:24:35)
- vishnoianil says he thinks we should take
vendor extensions, only if there's a strong committment to support
it (colindixon,
17:25:06)
- colindixon says that there are 3 main things
we're talking about (1) code organization into repositories, (2)
code organization in terms of the number of layers of abstraction
and it's implication for performance, (3) how we get people to
maintain the code (especially extensions) and if and how they get
shipped as part of a release (colindixon,
17:34:45)
- colindixon asks what people think about
aggregating all openflow projets under an openflow/ name (like
integration) (colindixon,
17:35:04)
- rovarga says he likes it, but for now, as long
as it's kept to the plugin, library, and extenions for now and maybe
expanding later based on our experiences (colindixon,
17:35:50)
- https://lists.opendaylight.org/pipermail/discuss/2016-May/006522.html
tykeal's comments on moving projects (colindixon,
17:36:14)
- vishnoianil says in his mind l2switch could
even join, rovarga says he worries that most projects are very
openflow specific and so everything might end up under
openflow/ (colindixon,
17:36:45)
- abhijitkumbhare agrees with rovarga to start
small (colindixon,
17:38:30)
- colindixon points out that tykeal says above
that moving projects would have to change their group IDs for
artifacts (colindixon,
17:38:54)
- rovarga says that as long as somebody goes
through and checks all the dependent projects and pushes the
patches, it should be pretty easy, the old artifacts will be around
for a while and it won't break anyone (colindixon,
17:41:53)
- colindixon asks under what circumstances (if
any) would we allow extensions into the openflow plugin project
itself? (colindixon,
17:43:05)
- abhijitkumbhare says that some crtieria might
be vendor neutrality and system tests (colindixon,
17:44:16)
- colindixon asks if we ship OF extensions that
are part Boron and then the project that was providing them (which
isn't openflowplugin) drops out of Carbon, do we think users will
correctly blame that project and not openflowplugin (colindixon,
17:45:51)
- colindixon's point is really that if we worry
about being held accountable for teams vanishing, merely keeping the
functionality out of openflowplugin itself might not help as much as
we'd like (colindixon,
17:47:26)
- it seels like we agreement on moving
openflow{plugin,java} and extensions under one group (colindixon,
17:49:35)
- vishnoianil asks if we should even ship vendor
extensiosn with the simultanteous release, e.g., just point to the
external jars/bundles/feature-repos intead of shipping with
it (colindixon,
17:50:33)
- rovarga notes that right now, projects can
provide extensions and be part of the release if they want to, we'd
need to change our rules to not allow them to participate
(colindixon,
17:52:38)
- colindixon and abhijitkumbhare say that good
examples on both sides of this are things like the Nicira
extensions, which we probably do want in OpenDayligth by default,
but there are other extensions which are clearly more niche and
might be find to not ship (colindixon,
17:56:18)
- rovarga suggests maybe having things need to be
standardized or implemented by two or more vendors first, might help
wiht that problem (colindixon,
17:57:17)
- colindixon worries that it might be hard to
make it clear that this was objective, also it doesn't help that we
also need to figure out how it will stay maintained (colindixon,
17:58:30)
Meeting ended at 17:58:53 UTC
(full logs).
Action items
- (none)
People present (lines said)
- colindixon (47)
- odl_meetbot (4)
Generated by MeetBot 0.1.4.