Following are the minutes of the unofficial telecon of the XRI TC at:
Date: Thursday, 24 September 2009 USA
Time: 2:00PM - 3:00PM Pacific Time (21:00-22:00 UTC)
ATTENDING
Eran Hammer-Lahav
Breno de Medeiros
Dirk Balfanz
Will Norris
John Bradley
Bob Morgan
Drummond Reed
Peter Davis
George Fletcher
Nat Sakimura
AGENDA
1) XRD 1.0 WD06 ISSUES
Again, the main focus of the telecon will be review of any final issues.
http://www.oasis-open.org/committees/download.php/34072/xrd-1.0-wd06.html
Several issues have been raised on the list on which we need final closure:
1A) Linked XRDs and #include vs. #replace semantics (last week's issue)
The #replace directive would delegate the entire XRD. Every other option involves some form of #include, and that introduces processing complexity we would rather avoid.
More limited forms of "include" can be defined by other specifications, and if one proves to be general enough, we can add that in a future version.
The strongest use case for delegation right now is at the host-meta level.
#DECISION: The consensus was to leave out XRD linking completely in 1.0 and let it be tacked in other specs that need it.
1B) Guidance in the spec on use of multiple <Rel> elements in a <Link>
Breno described the use case where one very general <Rel> like "describedBy" fully encompasses a more specific <Rel> (e.g., an "all-my-OpenID-metadata-is-here" <Rel>). He thinks that consumers need clear processing rules in this case (or providers need clear authoring rules).
Eran made the point that the inclusion of multiple Rels in a Link (regardless of the cardinality of URI or URITemplate elements in the Link) MUST be semantically be the same as if they were separated into multiple Links. For example, if you have a Link with two Rels and one URI, it can ALSO be expressed as two Links with one Rel each and same URI in each.
Breno made the proposal to force XRD providers fully parse out the combinations of Rels and URIs in links by changing Rel and URI cardinality to 1 or 0.
Will brought up that if they are semantically equivalent, the question is whether allowing 0-n Rels or URIs in a single Link actually provides any optimization. Breno is concerned that if we do so, it will lead XRD implementers to incorrect semantic understanding and thus incorrect implementations. Breno's 0-or-1 proposal is that we can prevent this by enforcing XRD providers to clearly state all the valid combinations in document order .
George made several observations about how this might affect performance based on his experience at AOL, but in the end they do not really affect this decision.
Drummond mentioned that this may produce a slightly more verbose XRD document, but John pointed out that it will compress to the same size.
This then led to the question of whether, if Rel and URI cardinality becomes 0-or-1, we should revisit these being attributes instead of child elements -- it certainly gives more weight to the arguments Dewitt Clinton was making about this in his feedback in August.
1C) Cardinality of <URI> and <URITemplate> in a <Link>
1D) <URI> and <URITemplate> elements: combining them, template
semantics, etc.
We did not have time to talk about these two issues - we need to persue them on the list.
2) SCHEDULE FOR COMMITTEE DRAFT 01 VOTE
It is still unclear when we can hold the vote - we need to close the issues above and publish a "Last Call" WD first. We will continue to move as fast as we can.
3) NEXT CALL
Next week at the regular time. Note that Drummond and John may not be able to attend due to more meetings they will be attending in Washington D.C.