14:00:16 <igordcard> #startmeeting network_common_flow_classifier
14:00:17 <openstack> Meeting started Tue Apr 18 14:00:16 2017 UTC and is due to finish in 60 minutes.  The chair is igordcard. Information about MeetBot at http://wiki.debian.org/MeetBot.
14:00:18 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
14:00:20 <openstack> The meeting name has been set to 'network_common_flow_classifier'
14:00:26 <davidsha> Hey
14:00:30 <igordcard> hey davidsha
14:00:31 <igordcard> hi all
14:00:55 <bcafarel> howdy
14:01:00 <igordcard> what's up bcafarel
14:01:39 <igordcard> agenda:
14:01:43 <igordcard> #link https://wiki.openstack.org/wiki/Neutron/CommonClassificationFramework#Discussion_Topic_18_April_2017
14:01:53 <reedip_> o/
14:02:00 <igordcard> hi reedip_
14:02:04 <davidsha> hey
14:02:20 <igordcard> let's dive in
14:02:25 <igordcard> #topic Closing the spec
14:02:48 <igordcard> a new spec was uploaded:
14:02:51 <igordcard> #link https://review.openstack.org/#/c/333993/14
14:03:14 <igordcard> but I've set W-1 recently to rework the database schema again
14:04:31 <igordcard> it has come to my attention that the AND/ORing of classifications isn't particularly easy for some basic scenarios, and that the fixed nature of the qos-inspired model makes it difficult to make that easy or make changes in the future
14:05:19 <igordcard> I'm interested in knowing more about what makes base/child classifications create problems when using OVO
14:06:06 <igordcard> see comment at line 236 of:
14:06:09 <igordcard> #link https://review.openstack.org/#/c/333993/13/specs/pike/common-classification-framework.rst
14:06:14 <davidsha> Ihar seems to be offline and he'd be the expert.
14:07:17 <igordcard> davidsha: yep probably the best is to reach out to him and ask for more details - I don't want us to commit to a model that will bring us deep trouble with OVO, but would like to have this flexible enough
14:07:56 <igordcard> apart from that, after the change regarding the schema (that will be up today), I believe the spec is essentially done
14:08:42 <davidsha> that's good.
14:10:29 <bcafarel> kind of same comment as the db, in REST api I did not see how to get the sub groups for a CG
14:10:59 <igordcard> bcafarel: there are no "sub groups", what do you expect them to be?
14:12:55 <bcafarel> igordcard: sorry I may still be catching up on these, the terminology says a CG is a grouping Classifications, or other Classification Groups
14:13:19 <bcafarel> so I was seeing it as visible in the group itself (as "sub groups")
14:13:38 <bcafarel> or is it just up to the service to implement the logic?
14:14:19 <igordcard> bcafarel: nice catch!
14:14:58 <igordcard> bcafarel: seems I wrote that while I still a different model in mind (which I will likely bring back on the next patchet)
14:15:10 <igordcard> * while I still had a
14:15:36 <igordcard> what I'm thinking is that groups can include other groups
14:15:47 <igordcard> that way, you can indeed mix and match ORs and ANDs
14:16:39 <igordcard> so you can say "match on this ip address, OR that ip address, AND make sure either is tcp port 80 traffic"
14:17:43 <igordcard> this would be an AND-group made of 1 classification and 1 OR-group
14:18:06 <igordcard> with the OR-group having 2 classifications inside
14:19:25 <bcafarel> ack, that was what I had in mind
14:19:35 <bcafarel> if that is for a later patchset no problem then :)
14:19:36 <igordcard> in terms of the database schema this should be neat and easy to model, but I have to investigate a bit more the API and OVO consequences
14:20:04 <igordcard> bcafarel: yeah I'll post something today
14:21:11 <igordcard> bcafarel: but the current patchset only has single-level grouping of either all-AND or all-OR
14:21:35 <igordcard> and if a CS wanted, it could consume multiple groups to create complex expressions
14:22:06 <igordcard> seems I don't have a topic for the PoC, so I will integrate it
14:22:12 <igordcard> davidsha: how's it going?
14:22:23 <igordcard> integrate it *here
14:23:14 <davidsha> igordcard: Going well, I'm just filling out the PoC at the moment. I'm starting looking into the classification groups as well but only very early stuff.
14:23:39 <davidsha> Less Pep8 fails for sure anyways :P
14:24:02 <igordcard> great
14:24:19 <igordcard> the classification groups is the core open debate right now
14:24:47 <igordcard> we start implemeting too early and we'll end up wasting effort if we can't get enough agreement on the spec
14:25:42 <igordcard> so please, all review the spec after the next patchset (due today) so we can more confidently implement a grouping model
14:26:01 <davidsha> ack
14:26:02 <igordcard> I'll talk to Ihar about the OVO issue
14:26:09 <reedip> ack
14:26:38 <igordcard> moving on
14:26:43 <igordcard> #topic Rights to the/a repo
14:26:52 <igordcard> for context:
14:26:58 <igordcard> #link https://bugs.launchpad.net/neutron/+bug/1476527
14:26:59 <openstack> Launchpad bug 1476527 in neutron "[RFE] Add common classifier resource" [Wishlist,Triaged] - Assigned to Igor Duarte Cardoso (igordcard)
14:28:01 <igordcard> so with agreement on the spec we can be given a repo and a subteam to work on and merge the code
14:29:34 <igordcard> it's probably going to be neutron/classifier and me + davidsha  as the initial team (spec and poc authors respectively), we can then grow the team as other dedicated contributors arrive
14:29:49 <davidsha> ack
14:31:34 <igordcard> while there's no clear agreement on the spec, I'll simply wait until any of the cores review the spec and decide to give us the repo/rights
14:31:46 <igordcard> once we have agreement I will actively ping them to set this up
14:31:53 <reedip> igordcard : subteam is a good idea
14:33:03 <igordcard> good good
14:33:32 <igordcard> alright, moving on..
14:33:43 <igordcard> #topic Open discussion
14:34:08 <igordcard> whatever you folks want to chat about
14:34:44 <bcafarel> except the weather
14:35:17 <davidsha> Naturally :P
14:35:22 <bcafarel> I still have to look at the poc (now that I am back up to date on the spec), sorry been busy elsewhere
14:35:22 <igordcard> where's free speech? :'(
14:35:36 <bcafarel> ok weather topics allowed then :)
14:36:11 <davidsha> I was hoping to have another PS for the classifier up by the end of the week, but It will probably be a bit off while I wait for the dust to settle around classification_groups.
14:37:03 <igordcard> davidsha: yeah... I might ping the most prominent voices around classification grouping after I submit my new PS
14:37:15 <davidsha> igordcard: ack
14:38:13 <igordcard> we'll never totally agree on the grouping, but I'm hoping we can at least have a solution with no major drawbacks
14:38:21 <davidsha> k
14:38:35 <igordcard> having said that my biggest concern is really the OVO issue at this point
14:40:02 <igordcard> bcafarel: any plans on turning networking-sfc into a CS?
14:41:26 <bcafarel> igordcard: climbing up the todo list, I'd like to get at least some minimal poc up soon
14:41:34 <igordcard> bcafarel: oh great!
14:41:37 <bcafarel> (when the unit tests stop breaking during the week-ends)
14:42:47 <igordcard> bcafarel: one of the things I forgot in the spec was to add ip/mac addr masks.. possibly there are other gaps there (when comparing to the Flow Classifier)
14:43:20 <igordcard> bcafarel: if I don't fix all the gaps, please leave a comment there
14:45:12 <bcafarel> igordcard: oh ok noted, I did not check thoroughly the details in the type tables
14:45:19 <igordcard> btw, stealing the topic from the fwaas meeting - who is going to the summit?
14:45:45 <igordcard> I wasn't originally going to, but am now
14:46:29 <davidsha> I'm not, good luck with the trip!
14:48:01 <bcafarel> I won't go either, we count on you to market the CCF concept!
14:49:37 <igordcard> :(
14:50:50 <bcafarel> I wonder how many engineers will turn up at the summit in the end (it is the first one after the PTG split)
14:50:56 <reedip> lol igordcard :)
14:51:15 <igordcard> are you reedip ?
14:51:23 <reedip> NOPE ...
14:51:36 <reedip> I have more dangerous missions to go through :)
14:58:41 <igordcard> alright dear contributors, let's finish the meeting
14:58:50 <igordcard> bye all
14:59:07 <igordcard> #endmeeting