See also: IRC log
2008-03-06 http://lists.w3.org/Archives/Public/public-sml/2008Mar/att-0048/20080306-sml-minutes.html
RESOLUTION: the 2008-03-06 minutes at http://lists.w3.org/Archives/Public/public-sml/2008Mar/att-0048/20080306-sml-minutes.html are approved.
Kumar has a proposal in comment #2.
No questions or concerns from WG members present.
No questions or concerns from WG members present on the proposal in comment #2.
Kumar: this is related to 5564 recursive definition of DerefExpr should be restored
Pratul: need to talk to MSM and John about whether the change suggested in 5564 forces us back to LC.
WG OK with change suggested in comment #2 of 5526.
RESOLUTION: WG approves change suggested in comment #1. Mark as editorial.
Ginny: need more time to review.
Pratul: put this on next week's agenda.
Pratul: the term "interchange set
validation" would also be affected.
... we had earlier discussion between "interchange set
validation" and "SML-IF model validation". We chose
"interchange set validation". Does that mean we prefer "set"
over "model"?
Ginny: I like the recommendation
to use "interchange model".
... this is different from calling it "SML-IF model".
Pratul: that would change "interchange set validation" to "interchange model validation".
Sandy: prefer "model" too, since we define "model" as "a set of documents".
Kumar: "interchange model validation" is not exactly the same as "SML model validation".
Kirk: right. "interchange set validation" includes "SML model validation".
Sandy: "interchange set validation" is a special invocation of "SML model validation", hence they should use the same term "model".
Kumar: I think the "interchange set" should include both the "model" and meta-data.
Kirk: I think "interchange set" only includes the documents. the meta-data is part of the IF document, not the interchange set.
Pratul: reading the definition of "interchange set". finding it somewhat confusing.
Kumar: again, maybe "interchange set" should include both the "model" and meta-data. IF is just the encoding format.
<pratul> This is what the spec says today
<pratul> The interchange set is the set of documents that constitute the SML model [SML 1.1] to be interchanged.
<pratul> Interchange set validation is the process of assessing the validity of the SML model [SML 1.1] represented by the interchange set while maintaining all assertions and interrelationships among the documents in the interchange set as defined by this specification.
Sandy: Kumar's suggestion is interesting to exploit. Feel that "set" should be used to refer to the "set of the documents"; "model" could be used to refer to more than the documents. That is, we may want to switch their places.
Kumar: different intuition.
"model" is defined by SML. Want to use "set" for what's
included in IF document.
... let's discuss offline, among the interested parties, and
come back with a proposal.
<scribe> ACTION: Kumar to discuss 5283 with interested parties, and come back with a proposal. [recorded in http://www.w3.org/2008/03/13-sml-minutes.html#action01]
<trackbot-ng> Created ACTION-176 - Discuss 5283 with interested parties, and come back with a proposal. [on Kumar Pandit - due 2008-03-20].
Kirk: part of 5341. suggest to hold it off until we discuss EPR note.
Pratul: should it be a duplicate of 5341?
Kirk: should keep it separate.
Kumar: already fixed in LC draft.
Pratul: put a comment in the bug
explaining the current situation in LC.
... maybe we want to have a new keyword for issues that are
waiting for the commentator to respond.
Ginny: the keyword "decided" seems what we need.
Pratul: change bugs in this category to use keyword "decided".
RESOLUTION: adopt the suggestion in 5528 to remove the xs:import.
Pratul: (clarify Henry's comment) the schema document in C has no SML reference constraint, which means having sml references (or not) in the instance doesn't affect model/document validity.
Kumar: if the model only contains these documents, then Henry is correct.
Pratul: look at comment #1. even when there is no reference constraint, there are still things that needs to be checked for references. these may affect model validity too.
Kumar: what's in C is not wrong. we may want to add a note to address his concern.
Sandy: do we have examples in Appendices that show how target* constraints work?
Pratul: may want to open a separate issue for it.
RESOLUTION: mark 5529 editorial. Fix it in the spirit of comment #1. WG needs to review the proposal.
Kumar: suspect Henry's referring
to an earlier draft, because of the mention of xmlns().
... suggest to add a comment asking for clarification from
Henry.
Pratul: suggest editors' draft should point to the current version; editors define a checklist internally, requiring that the drafts be updated to point to the correct "previous version" before publication. (basically MSM's points 1 & 2.)
All: agree.
RESOLUTION: mark 5532 editorial, by adopting MSM's suggestions in the bug report.
Sandy: I think we discussed whether to honor schema defaulted sml:ref before and decided to allow that to be implementation dependent or configurable in processors. this should address Henry's concern, at least in the core spec.
Kumar: also remember that discussion from one of the F2F meetings. can't find the corresponding text.
Pratul: I think the main scenario where you want to use schema defaulted sml:ref is when you receive an IF document.
Sandy: suggest to update section
4.1.1, to allow both the original infoset and the PSVI to be
used for reference recognition.
... the current paragraph starting with 'this mechanism"
becomes a reason for why allowing to look at the original
infoset could be useful. we could add another sentence to
explain why using PSVI could be useful: to use default
sml:ref=true.
Pratul: sounds reasonable. there are scenarios for both schema-less recognition, and schema-defaulted recognition.
Ginny: that means it can be done in either way?
Kumar: should it be "use schema default if performing validation, otherwise either way"?
Ginny: given that SML only defines validator behavior, do we need to be worried about the schema-less scenario?
<pratul> Schemaless identification of references was driven by scenarios that required references to be quickly identified
<pratul> when models were being exchanged, stored, or retrieved and these operations required references to be transformed
Pratul: explains why schema-less identification is useful. see Pratul's comment above.
Ginny: so this is convenience; it doesn't affect validation.
<scribe> ACTION: Sandy to work on a proposal to address bug 5541. [recorded in http://www.w3.org/2008/03/13-sml-minutes.html#action02]
<trackbot-ng> Created ACTION-177 - Work on a proposal to address bug 5541. [on Sandy Gao - due 2008-03-20].
Sandy: I think we intentionally
left this open. processors may choose to use xml:base
attribute, or other ways to establish the base. SML doesn't
specify a fixed behavior.
... we answer this question in IF.
Pratul: the only thing we talk about is when you have a fragment, then it's pointing to the same document.
Ginny: then anything we need to do for this bug?
Sandy: it may help to have an explicit statement about the impl-dependent nature of base uri.
Ginny: we already have it. in section 4.3.1, bullet 2.a.
Sandy: section 4.3, bullet 3.b is the general requirement for schemes using target-complete identifiers; section 4.3.1, bullet 2.a is how the URI scheme is handled.
RESOLUTION: resolve 5542 as invalid. see sections 4.3 and 4.3.1.
Last Scribe Date Member Name Regrets pending 2007-08-30 Lipton, Paul until mid-January 2008 2008-01-22 Wilson, Kirk 2008-01-22 Eckert, Zulah 2008-01-23 Smith, Virginia 2008-02-07 Lynn, James 2008-02-14 McCarthy, Julia 2008-02-21 Kumar, Pandit 2008-03-06 Boucher, Jordan 2008-03-13 Gao, Sandy Exempt Arwe, John Exempt Dublish, Pratul Exempt MSM Exempt PH