#opendaylight-group-policy: ODL-GBP-ARCH

Meeting summary

1. 1. we are talking about need to reduce # of contracts per subject (dconde, 18:20:53)
   2. but what about based on flow? (dconde, 18:21:01)
   3. that's doable with conditions! - says dvorkinista (dconde, 18:21:09)
   4. if it's anomalous we can apply those to sites provided by said servers. (dconde, 18:21:27)
   5. but mickey says" what is the def iof flow" (dconde, 18:21:39)
   6. sanjay says "user X comm to wiki site A on HTTP" that flow is what Sanjay wants to apply security, improve experience, etc. but you want to do something different. (dconde, 18:22:10)
   7. dvorkinista says you can mark them (dconde, 18:22:21)
   8. but sanjay says that only subset needs to be set. classifiers vs. conditions is unclear (dconde, 18:22:41)
   9. filters are HTTP -- only difference is actions says dvorkinista (dconde, 18:22:53)
   10. sanjays says what about on per-site (dconde, 18:23:26)
   11. but if we have different sites, it's a different EPG -- says dvorkinista (dconde, 18:23:39)
   12. info we can cover that in a clause. (dconde, 18:24:07)
   13. functional equivalent thing are doable. not do it instance specific. (dconde, 18:24:23)
   14. we have a functonal way to achive that in current model with clauses, etc. (dconde, 18:25:23)
   15. we are worried about proliferation of subjects - says Sanjar (dconde, 18:29:08)
   16. dvorkinista says we can intro containment within subject (dconde, 18:29:18)
   17. same classifiers and multiple ACTION sets (dconde, 18:29:32)
   18. in the world of controls -- it's never in run time. (dconde, 18:31:45)
   19. there is an understanding of expected actions. nothing ins unplanned, says dvorkinista (dconde, 18:32:13)
   20. this is in response to conditional labels or some other exception mechanism. (dconde, 18:32:26)
   21. sajnay asks if we have lots of flows…then we….. (dconde, 18:33:00)
   22. dvorkin says -- we can do anomaly detection per flow (dconde, 18:33:13)
   23. anjays wants to see how to get away from too many subjects per flow (dconde, 18:34:55)
   24. dvorkinista says it is not per flow. (dconde, 18:35:15)
   25. we want to pre plan conditions to react consistently. (dconde, 18:35:31)
   26. dvorkinista parts of contracts can be mutated over time (dconde, 18:36:14)
   27. lenrow says if this is best practice for expected events -- that's OK (dconde, 18:36:26)
   28. dvorkinista says let's figure out planned stuff vs. dynamic stuff. (dconde, 18:36:59)
   29. do we want to extend concept of a session and piggy back on top of that. (dconde, 18:37:38)
   30. sanjay asks how do we structure flows to KNOWN policies? (dconde, 18:38:50)
   31. dvorkinista semanticlly we can express is as for two EPs, each belong to an EPG, in context of a given contract, we need to overwrite a set of subjects. that's an exception. (dconde, 18:39:32)
   32. we want to avoid a backdoor into this. (dconde, 18:39:53)
   33. it's OK to have exception case (like using Asmbly for dev drivers) (dconde, 18:40:31)
   34. in the third option in yesterday's meeting, only thing we change are that there are not rules, but we put into filters of a subject. (dconde, 18:41:12)
   35. mickey -- how do we over ride or how to figure dyn add conditions, etc. (dconde, 18:44:56)
   36. dconde asked whether we want these conditions to be as orthogonal a possible. (dconde, 18:50:41)
   37. dvorkinista says yes (dconde, 18:50:47)
   38. the conditions are planted by app owners (dconde, 18:50:55)
   39. some form of melt-down that needs an exception to deal with anomaly. (dconde, 18:51:11)
   40. we cannot alter intent (dconde, 18:51:38)
   41. we want to deal with exceptons without changing intent. (dconde, 18:51:55)

Action items

1. (none)

People present (lines said)

1. dconde (47)
2. odl_meetbot (4)
3. alagalah (2)
4. regXboi (0)