Chat transcript from room: tosca
2012-11-05 GMT-08:00
[07:58] Room information was updated by: Matt Rutkowski (IBM)
[07:58] Please record your attendance: https://www.oasis-open.org/apps/org/workgroup/tosca-interop/event.php?event_id=33596 Audio Conference Access Numbers USA: 1-888-426-6840, 215-861-6239 Passcode: 627 057 12 # IBM SmartCloud Meetings Conference ID: 6789260 External URL: https://www.ibm.com/collaboration/meeting/join?schedid=6789260
[08:06] Paul Lipton (Co-Chair) private: Sorry, I am delayed. Am trying to dial in now
[08:10] Matt Rutkowski (IBM): move approve to approve last week’s minutes, Paul 2nds, no objs.
[08:16] Matt Rutkowski (IBM): Tobias: re name property, some discussion was had several months ago
[08:16] Matt Rutkowski (IBM): Tobias: when this came up, as soon DSL are supported, the id and name fields would appear redundant
[08:17] Matt Rutkowski (IBM): Paul suggests that we find the issue we discussed this on.
[08:17] Matt Rutkowski (IBM): Ken: recalls discussion of xlation to other formats like JSON
[08:17] Matt Rutkowski (IBM): Tobias: TOSCA 20 issue 6
[08:18] Matt Rutkowski (IBM): Paul: subtask TOSCA 21, had 3 poss. solutions
[08:18] Matt Rutkowski (IBM): Dale: gen. Q about referential apparatus
[08:19] Matt Rutkowski (IBM): Dale: has been trying to process instance and build up data structs.
[08:19] Matt Rutkowski (IBM): Dale: there are 2 ways things are related in an instance, the type attribute in its QNAME style reference, its type reference
[08:19] Matt Rutkowski (IBM): Dale: this is used to pickup / inherit type properties
[08:20] Matt Rutkowski (IBM): Dale: in other case the IDREF is used and the name is window dressing
[08:20] Matt Rutkowski (IBM): Dale: the att. ref on source.
[08:21] Matt Rutkowski (IBM): Dale: we should say more about these ref. mechanisms
[08:21] Matt Rutkowski (IBM): Dale: use of names is tricky
[08:22] Matt Rutkowski (IBM): Thomas: re: Dale, QNAMES and IDREF is std. XML processing mechs.
[08:22] Matt Rutkowski (IBM): Thomas: the rule is that everything (elements) in same XLM doc. uses IDREF
[08:23] Matt Rutkowski (IBM): Thomas: QNAMES permit external refs
[08:23] Matt Rutkowski (IBM): Thomas: name is kind of used for navigation for reqs/caps
[08:23] Matt Rutkowski (IBM): Thomas: because we name rels in node types (named slots0
[08:23] Matt Rutkowski (IBM): Thomas: want to navigate node from a model to another node
[08:23] Matt Rutkowski (IBM): Thomas: a 3rd way of ref. nodes
[08:24] Matt Rutkowski (IBM): Dale: agree QNAMES/IDREF uses, a 3rd mech. seems confusing
[08:25] Matt Rutkowski (IBM): Dale: could be used for matching, but the names do nat match verbatim from sample
[08:25] Matt Rutkowski (IBM): Dale: what has been documented?
[08:26] Matt Rutkowski (IBM): Dale: what do i rely upon building a relational structure, how does a req fulfill a cap.?
[08:27] Matt Rutkowski (IBM): Thomas: using the IDREF mech, is the correct thing to do
[08:27] Matt Rutkowski (IBM): Thomas: via source/target should work
[08:27] Matt Rutkowski (IBM): Thomas: is a more advanced use case, Derek can you illustrate?
[08:28] Matt Rutkowski (IBM): Tobias: That is use case he cares about, everything turns into a platform, i want an apache web server (later than some vers)
[08:28] Matt Rutkowski (IBM): Tobias: falls out to exp. of reqs. by some symbolic
[08:29] Matt Rutkowski (IBM): Dale: what methods do i need to make those matches? Partial string matches?
[08:29] Matt Rutkowski (IBM): Tobias: we would use our own strucs off a node type with direct string match and less than/greater than or equal matching
[08:30] Matt Rutkowski (IBM): Ken: proposed some operations some months ago
[08:32] Matt Rutkowski (IBM): Matt: 3rd option is fuzzy still (QNAMES and IDREF are understood
[08:33] Matt Rutkowski (IBM): Dale: a matching process, not navigation necessarily
[08:33] Matt Rutkowski (IBM): Tobias: navigation enters rapidly; OS for example.
[08:33] Matt Rutkowski (IBM): Tobias: need to know how to affect the OS via some nav. relationship
[08:34] Matt Rutkowski (IBM): Dale: once we pull out source/target and ref attrs. it gives us the nodes on rel.
[08:34] Matt Rutkowski (IBM): Dale: need to then look at req./cap list
[08:34] Matt Rutkowski (IBM): Tobias: always need to resolve navigation as well
[08:35] Matt Rutkowski (IBM): Thomas: only using QNAME and IDREF; defer to Derek
[08:38] Matt Rutkowski (IBM): Derek: type names need to be fixed
[08:39] Matt Rutkowski (IBM): Derek: instance names, node templates can be more lax
[08:39] Matt Rutkowski (IBM): Derek: make a list of questions.
[08:40] Matt Rutkowski (IBM): Dale: QNAMES ref back to type is not an issue; use of ID and IDREF to ref. nodes within a relation template are not issues
[08:41] Matt Rutkowski (IBM): Dale: but, are relational strucs being built up by matching names; req/cap matching, do we build up a list based upon names?
[08:41] Matt Rutkowski (IBM): Derek: that is type matching, matching by class. if types have qnames can use that
[08:41] Matt Rutkowski (IBM): Dale: use of fields within; its less clear, what is matching procedure?
[08:42] Matt Rutkowski (IBM): Derek: its not specified, only type matching
[08:42] Matt Rutkowski (IBM): Derek: can match subclasses (candidate matches)
[08:43] Matt Rutkowski (IBM): Dale: intent of spec. is that cap matching s done by type or subtype
[08:43] Matt Rutkowski (IBM): Derek: correct, can differentiate by many things which are not specified, type matching should be declared
[08:44] Matt Rutkowski (IBM): Dale: could be a primer issue (illustration, with type/subtypes)
[08:44] Matt Rutkowski (IBM): Derek: absolutely, any impl. would encounter this
[08:44] Matt Rutkowski (IBM): Matt: we perhaps do not have use case in primer that permits us discussion of this
[08:44] Matt Rutkowski (IBM): Dale: this critical
[08:45] Matt Rutkowski (IBM): Paul: is this a spec issue as well?
[08:45] Matt Rutkowski (IBM): Derek: It perhaps needs to be clear in spec. of course should be in primer
[08:45] Matt Rutkowski (IBM): Derek: need an exp. language to resolve ambiguities
[08:46] Matt Rutkowski (IBM): Matt: heard Tobias and Ken are going forward with own exp. language
[08:46] Matt Rutkowski (IBM): Is this a long term interop problem
[08:46] Matt Rutkowski (IBM): Paul: Primer is place we agreed to do this
[08:46] Matt Rutkowski (IBM): Matt: will pass this to primer
[08:48] Matt Rutkowski (IBM): Paul: create a vnext JIRA issue? Asks Dale to open. Dale agreed.
[08:50] Matt Rutkowski (IBM): Tobias: all these matching ops. do not emerge until u have two separate roles, node types for reuse (provide) and separate consumer
[08:50] Matt Rutkowski (IBM): Tobias: we do not need this perhaps for SugarCRM interop
[08:50] Matt Rutkowski (IBM): Tobias: as we know node types
[08:50] Matt Rutkowski (IBM): Tobias: we agreed at TC not to pursue this
[08:51] Matt Rutkowski (IBM): Tobias: like Linux distros, packages matching is done local to distro
[08:51] Matt Rutkowski (IBM): Tobias: perhaps match by URL with version attr at end of the url
[08:52] Matt Rutkowski (IBM): Tobias: these were past suggs, these roles were not discussed in text(?)
[08:56] Matt Rutkowski (IBM): Matt moves to adjourn, Dale 2nds, no objs.