================================= #opendaylight-group-policy: MODEL ================================= Meeting started by dconde at 17:05:16 UTC. The full logs are available at http://meetings.opendaylight.org/opendaylight-group-policy/2014/model/opendaylight-group-policy-model.2014-05-02-17.05.log.html . Meeting summary --------------- * jmedved was to convert model to YANG (dconde, 17:05:42) * and workarounds and changes to model (dconde, 17:05:50) * recap coming w/ requirements of model (dconde, 17:06:11) * YANG cannot expressed self contained constructs (dconde, 17:06:23) * Discussion of YANG and UML capabailities (alagalah, 17:06:33) * linear chains are preferred to mimic containment (dconde, 17:06:36) * jmedved was to convert model to YANG (alagalah, 17:06:42) * recap coming w/ requirements of model (alagalah, 17:06:51) * YANG cannot expressed self contained constructs (alagalah, 17:06:55) * how to implement nested logical stmts (dconde, 17:07:04) * we can either keep formulas as blobs or str (dconde, 17:07:20) * specifically for matchers (alagalah, 17:07:24) * to simplify the matchers (dconde, 17:07:47) * we will use YANG as a authoritative description of model, rather than diagram (dconde, 17:09:00) * from readams (dconde, 17:09:08) * ought to extend YANG? (dconde, 17:09:29) * ACTION: dvorkinista (Mike D) to modify Matcher logic to be either ANDs, ORs, with Label EXCLUDE for negation. This will be a single level ie (A AND B AND C), (A OR B OR C), EXCLUDE (A AND B) (alagalah, 17:09:29) * but that is second phase. (dconde, 17:09:43) * dvorkinista recommends no circumstances and scoring. (dconde, 17:10:17) * assume there exists a governance system to update the endpoints. (dconde, 17:10:29) * as an external input (dconde, 17:10:38) * here (mickey_spiegel, 17:10:48) * it will not reduce the attractiveness of the model. (dconde, 17:11:14) * lenrow says we don't need to do all the use-cases. (dconde, 17:11:47) * no matter what we do, we have to architect for extensibility. (dconde, 17:12:27) * MD-SAL / YANG requirements discussion - broader discussion (alagalah, 17:12:41) * best thing for community is to look at readams's paper on how we can achieve goals for simplification. (dconde, 17:13:33) * paper to be finalized and shared next week. (dconde, 17:13:46) * we will assume jmedved is in agreement. since he wasn't on this moment. (dconde, 17:14:37) * we can look at what readams has transcribed. (dconde, 17:15:37) * review of readams's yang model for 0.97 (dconde, 17:15:59) * we are reviewing the doc now on hangout screen share now. (dconde, 17:16:35) * early next week (dconde, 17:17:41) * trying to get inhr and obj modl to map to UML doc. (dconde, 17:18:14) * challenge. model is complex (dconde, 17:18:25) * starting w/ jmedved's model and adding constructs - issues are complexity and decipher the "runes" w.r.t. inconsistencies that readams will rsolve w/ dvorkinista (dconde, 17:19:29) * same type @ two levels of hier. (dconde, 17:19:42) * he is using grouping but still needs to be mapped to a different location. (dconde, 17:20:26) * running into restrictions Java that is generated. (dconde, 17:22:35) * other challenge. relator object (dconde, 17:23:15) * inheritance is problematic. children override values, so what should be in base class? (dconde, 17:23:52) * are you using AUGMENT? readams is using groupings. (dconde, 17:24:21) * discussing whether we are overriding base types. how best to do that? (dconde, 17:25:21) * if super class has restrictions, we can reason at that level? or shall we use REFINE stmt? It's all tricky. (dconde, 17:26:22) * dvorkinista says we used to have match type in relator in original. so we have a vestigial item that needs to resolved. (dconde, 17:27:23) * it's possible to just have a relator. use a target, selector, and that's it. can be named or not-named. (dconde, 17:28:01) * dvorkinista is fine not using inheritance. (dconde, 17:29:52) * relator can be a providing selector or a consumer selector. do not treat them as separate things. (dconde, 17:30:15) * once we have a proper model, it will be resolved in a concrete yang model (dconde, 17:32:33) * yang becomes authoritative. (dconde, 17:32:51) * but let us remembe that UML is easier to read... (dconde, 17:33:36) * ACTION: alagalah will write a whitepaper that describes the UML model. (dconde, 17:34:19) * and text is ambiguous. so we do not want to replace UML (dconde, 17:35:01) * should we annotate UML? dvorkinista prefers that. (dconde, 17:35:36) * we will disucss how best to do it (dconde, 17:36:20) * ACTION: we will annotate UML instead -- alagalah (dconde, 17:36:32) * lenrow has a "for dummes" draft in progress. (dconde, 17:37:07) * open forum. (dconde, 17:38:25) * mickey_spiegel notices that many pages of model refers to same thing. dvorkinista says it is used to describe concept. (dconde, 17:39:09) * alagalah says each page tries to express a concept in context. (dconde, 17:39:26) * ACTION: dvorkinista to restruct model in structure, definition use (dconde, 17:40:24) * lenrow questions. (dconde, 17:40:47) * is traffic chaining expression difficult? (dconde, 17:41:19) * we are missing some items. like sensitivities ! (dconde, 17:41:41) * dvorkinista will update model according to how it was expressed in YANG. (dconde, 17:42:09) * we meant to say service chaining, not traffic chaining. (dconde, 17:42:44) * let us try to model this. we are trying to pull in people from L4-L7 companies for their perspective. (dconde, 17:43:29) * let's have a discussion on what to include -- (dconde, 17:44:39) * we are speculating what RADware wants to do, or F5. (dconde, 17:45:12) * there is a project proposal for service chaining and see if it's common or not. (dconde, 17:45:23) * new svc chaining encapsulation type -- being proposed. (dconde, 17:46:35) * LINK: https://wiki.opendaylight.org/view/Project_Proposals:Service_function_chaining (raghu67, 17:47:04) * project is service function chaning that is being proposed. (dconde, 17:47:30) * visions from ETSI and IETF, etc. (dconde, 17:47:38) * there are lots of projects with some overlap (dconde, 17:48:43) * we can render to it. (dconde, 17:48:53) * mickey_spiegel says - is this a matter of rendering, or referring to their model? (dconde, 17:49:56) * they are concentrating on instance. GBP can focus on policy. this is complementary. (dconde, 17:51:32) * we will have a conversation. (dconde, 17:52:11) * ACTION: alagalah will talk w/ ewarnicke (dconde, 17:52:35) * ACTION: alagalah will talk w/ ed warnicke regarding service chaining. (dconde, 17:53:03) * their diagram has more stuff than svc chaining ought to have. (dconde, 17:55:29) Meeting ended at 17:56:19 UTC. Action items, by person ----------------------- * alagalah * alagalah will write a whitepaper that describes the UML model. * we will annotate UML instead -- alagalah * alagalah will talk w/ ewarnicke * alagalah will talk w/ ed warnicke regarding service chaining. * **UNASSIGNED** * dvorkinista (Mike D) to modify Matcher logic to be either ANDs, ORs, with Label EXCLUDE for negation. This will be a single level ie (A AND B AND C), (A OR B OR C), EXCLUDE (A AND B) * dvorkinista to restruct model in structure, definition use People present (lines said) --------------------------- * dconde (82) * alagalah (10) * odl_meetbot (5) * lenrow (3) * raghu67 (2) * jmedved (1) * mickey_spiegel (1) Generated by `MeetBot`_ 0.1.4