#opendaylight-meeting: md-sal interest call

Meeting summary

1. jersey migration (colindixon, 16:44:07)
   1. rgoulding_ is still testing things and he hopes it will be ready by next week, it's a pain to test things becuase autorelease is broken, and we can't test version bumps without building the world (colindixon, 16:45:14)
   
   
2. clustering behavior (colindixon, 16:45:19)
   1. there's a *long* discussion about the semantics of different events and how we should tell apps to handle things in the MD-SAL (colindixon, 16:46:21)
   2. this ends with moiz and colindixon agreeing that they should write up a clustering best practices document (colindixon, 16:47:06)
   
   
3. default values in YANG (colindixon, 16:47:12)
   1. ssaxena asks are we supporting default values in various YANG constructs (colindixon, 16:47:36)
   2. rovarga says that the spec says that "the system should behave as though the default value exists when unspecified", but not that the node actually has to exist (colindixon, 16:49:41)
   3. https://tools.ietf.org/html/rfc6020#section-7.6.1 (colindixon, 16:50:29)
   4. rovarga says the current appraoch is that applications to provide that property on their own (colindixon, 16:52:35)
   5. in essence it is the programmer's responsiblity to provide the appropriate behavior based on their out-of-band knowledge of the model (colindixon, 16:53:46)
   6. to help this, there are recent patches that provide default values for binding-aware DTOs on reads and data change notifications (colindixon, 16:55:18)
   7. rovarga notes that his reading of the spec is that the default nodes should not be returned as existing, only the behavior should change based on default values (colindixon, 16:56:25)
   8. ttkacik says that RFC 6243 spells out that there are particular functions which allow for reads with explicit rendering of the default nodes and thus implies that normal reads shoudl not render it (colindixon, 17:04:30)
   9. there is a broader issue around how to both respect the YANG spec to derive the maximum value from it, but not suffer too much from places where it makes thins counterintuitive for our developers and users (colindixon, 17:06:22)

Action items

1. (none)

People present (lines said)

1. colindixon (17)
2. odl_meetbot (3)