W3C SML F2F -- 23 Jan 2008

<scribe> scribe: Virginia Smith

<scribe> scribenick: ginny

schemaComplete discussion for Bug http://www.w3.org/Bugs/Public/show_bug.cgi?id=5395/

MSM proposal for 5395 schemacomplete wording is at http://lists.w3.org/Archives/Public/public-sml/2008Jan/0098.html

John: Version A/B correspond to option 1 from yesterday's discussion, Version C corresponds to option 2

Pratul: SML-IF model validation is not defined. What is an SMl-IF model and why do we need this term?

John: SML-IF model is the SML model represented by the interchange set

Sandy: we agreed to define this new term; you can say a model is embedded in the SML-IF doc
... we want aliases to participate in the model validation

Pratul: fine to say that if you are validating a model represented in an IF doc, here are the requirements for that.

Kumar: sml-if model validation is 2 steps: unpacking and then validating

Pratul: just talk about first step. we don't need to talk about validation since this is covered in SML spec

Sandy: can't separate the steps because some of the requirements must be performed during validation

Kumar: SML model validation assumes schema set already constructed so that step can be separate

<scribe> ACTION: Pratul to append comment to Bug #5395 concerning the term "SML-IF model validation" [recorded in http://www.w3.org/2008/01/23-sml-minutes.html#action01]

<trackbot-ng> Created ACTION-165 - Append comment to Bug #5395 concerning the term \"SML-IF model validation\" [on Pratul Dublish - due 2008-01-30].

Ginny: should schema be plural in 5.2.3 proposal?

Sandy: yes

Ginny: should B/C say schema... "MUST contain"

John: no, these proposals are declarative. The text as is is ok for normative text.

<johnarwe> MSM said his intent was: 5.2.3 versions A and B were equivalent in semantic intent. A is in the imperative or behavioral style, B is in the declarative style (what is true by definition). Both of them correspond to yesterday's item 1, where the behavior of a conforming SMLIF validating consumer is constrained and NO warrant from the producer is implied.

Zulah: version B is not consistent. 2nd sentence is behavioral.

John: agreed, need to fix this.

Kumar: what about built-in schema? this is outside interchange set

John: MSM said that, since they are built-in, they are not "retrieved from outside"

<johnarwe> 5.2.3 version C is in the declarative style, and corresponds to yesterday's option 2 (same constraint as yesterday's option 1, AND a warrant from the producer that no "element missing" or "attribute missing" validation constraints would be violated)

Sandy: it should say "schema document" rather than "schema component" in the proposed text.

<johnarwe> In the non-normative notes that follow, alternative 1 corresponds to either version A or B and alternative 2 corresponds to version C. I.e., if the wg selects either version A or version B conceptually, then alternative 1 would be selected as well. All of these may need further refinement to match the stated intent and/or editorial changes, MSM did not consider them final text.

John: previous consensus was that we wanted option 1 at a minimum

Kumar: ok with either A/B or C

Sandy: prefer B

ginny: why not C?

Sandy: implementation of C requires reference to error codes; may be difficult to get these error codes and some processors will provide this and some not.
... prefer B over A because it says something about the intent of the producer which is the attraction of C in the first place

John: Proposal is to select A/B not C. Any objections?

No objections heard.

John: Proposal on the table for choosing B over A

Valentina: prefer A; B does not state that "all schemas are included"

Sandy: A says do not go outside, B says, given the resulting schema - nothing came from outside the interchange set

Valentina: what if something is missing when you apply B?

John: ok, then model is invalid. B does not state that IF contains everything you need
... Consider B with Alternative 1 notes
... Any objections to the statement that B along with Alternative 1 notes expresses our intent?

Valentina: prefer A

Zulah: spec is written like A not B; prefer B but asking editors if this inconsistency is a problem

John: we agreed to review the spec and remove the behavioral statements
... Proposal B with Alternative 1 plus careful editorial attention

Kirk: I have a reservation until I see editorial work.

No other objections heard.

Resolution: Proposal B with Alternative 1 plus careful editorial attention. Editors to mark as needsReview when complete.

Attendees: John, Pratul, Kumar, Ginny, Kirk, Sandy, Zulah, Valentina

Kirk: The following action item is from yesterday but was not entered correctly.

<Kirk> ACTION: Virginia to open bug on how we should consider Schematron [recorded in http://www.w3.org/2008/01/23-sml-minutes.html#action02]

<trackbot-ng> Created ACTION-166 - Open bug on how we should consider Schematron [on Virginia Smith - due 2008-01-30].

Review of SML with Schema WG

<Kirk> John: Announced that the CG group has committed to Zurich as the site of the June mtg. It is up to us to decide whether we will be going.

Kumar is making a slide presentation

<ht> No, I think not, but a pointer to the slides would be nice

John: SML does not say anything about conflicting schemas

<MSM> slide 4 has been changed to address floor and ceiling issues

<MSM> [slide 5, substantially the same]

Kumar: Schematron is used as is, no extensions. Schema is extended.

John: SML defines rule bindings (for Schematron rules) but does not specify how to do that. SML-IF does specify how to interchange these rule bindings.
... these bindings can be expressed using wildcards
... SML primarily defines concepts useful across domains. SML-IF defines concrete syntax for exchanging models
... original context is system management. Recognition in WG that this can be applied more generally. Haven't formally recorded use cases.

<MSM> [if we are going to want a bat phone for discussion after the presentation, i can go fetch the bat phone. Are we going to want it?]

<MSM> [the plan is to register an xpath1() XPointer scheme and use it; fragment identifiers use XPath 1.0 plus deref() function

John: cross document reference 'schemes' can be defined by user; SML defines only 1 - the SML URI scheme
... how you recognize the scheme is specified in the *definition* of the reference scheme.

<ht> HST is happy with http://dev.w3.org/cvsweb/~checkout~/2007/xml/sml/build/sml.html?content-type=text/html;%20charset=utf-8#SMLXPath1_Scheme

<ht> No discussion necessary

MSM: if you don't recognize them, you ignore them

Noah: 2 instance documents and I want to express some constraints involving both documents
... trying to understand basic cross-references; do they show up in instances or schemas

MSM: these concepts (validation rules) show up in schemas

<ht> HST observes in passing that point 4. in section 4.3.1.1 thereof is open to misinterpretation, and suggests that "the containing element" be changed to "the parent sml:uri element"

<ht> HST is mildly surprised that barename references are not supported

MSM: can express key/keyref constraints if the targets are in the same document. you can use sml:key/keyref to do this across documents

MSM is explaining that SML references appear in instance documents

Noah: TAG is sensitive to new ways to reference things on the web that are not URIs.

John: the content of the SML URI reference scheme *is* a URI

Noah suggests that SML WG send note to TAG describing this.

ACTION: Working group to email TAG explaining that SML uses URIs to refer to web docs [opened by John Arwe on 2008-01-30]

<MSM> [note that the provision of alternative pointing mechanisms can be used to allow systems working locally to use end point references, if they choose to.]

<Jonathan> there are already several ways to do inter-document references, i assume the payoff here is the ability to use such references in validation, is that true?

<Jonathan> and what is the motivation for using markup for SML references?

<MSM> Yes, there's payoff in being able to validate the references, and to use the references for validation. One motivation for specifying specific markup is to allow systems to recognize SML references reliably.

<Jonathan> i see a syntax that my xpath parser would parse - is that true xpath, a subset, or what?

<MSM> See the uri above for details. Essentially, it's an xpath 1.0 location path, extended with the deref() function in the dynamic function space, and possibly with a few constraints on the expression

<johnarwe> full xpath 1.0. you can find this in section 4.3.1.1 of the SML (not SMLIF) editor's draft

<Jonathan> so what am i supposed to do with these declarations? is it defined purely on the schema level for validation, or are other operations on these cross document definitions supported?

<MSM> (Deref calls must come first, for example)

<johnarwe> which decls, specifically

<Jonathan> cross-document references

<MSM> [which declarations?]

<MSM> the key declarations in the example(s) are in schema documents

<MSM> and used for validation

Kumar: you can nest deref() functions

<MSM> if you believe a set of documents to be valid against a schema with such sml:key definitions, you can readily use deref() to dereference pointers and find things in other elements

<MSM> Strictly speaking, the idea of SML references and the ideas of sml:key/keyref and unique constraints are loosely coupled not tightly coupled (you can use an sml reference whether there's an sml:key constraint to guarantee its safety or not). But in practice, often deref() calls in SML model instance documents will be calls with XPaths that are guaranteed by the sml:key constraints to return (the right) things

<Jonathan> i guess i'm asking this: is referential integrity the main reason for sml?

<MSM> At a first approximation, I think the answer is yes.

<MSM> Certainly when I look at SML from an XSDL point of view, what I see is first and foremost "things XSDL might have provided for referential integrity checking, but did not", supplied as compatible extensions.

<MSM> At a secondary level, there are also provisions in SML-IF for more reliable definition of things like the binding of schemas to instances

<Jonathan> it looks like the submission was based on a subset of xml schema, but the current draft uses all of xml schema and a profile of schematron, right?

Sandy: there is some overlap between schema 1.1 and SML (Schematron) but they are for different use cases

MSM: Schema 1.1 cannot replace Schematron completely

David: suggests using a different example in the slides since that example can be covered in schema 1.1

<johnarwe> as part of the initial wg discussions, we decided to lift some of the restrictions made in the submission, eg wrt xs:redefine. the restrictions dropped were believed to be specific to certain contextual usages, so dropping them allows sml to be used more widely.

Kumar: in SML you can author constraints and dynamically bind them to different instance documents

<Jonathan> johnarwe - did you drop all such xml schema restrictions, or only some of them?

<johnarwe> usage of sml w/in the specific contexts the submission had in mind might require those same contextual restrictions on how SML would be used, but those restrictions are part of the contextual usage of SML not a basic feature of SML itself.

<Jonathan> gotcha, that makes sense

<johnarwe> I believe we removed all such restrictions.

<Jonathan> great - thanks

<Jonathan> ( trying to make sure i'm not looking at a superset of a subset design )

<johnarwe> fwiw, we did find during this week's mtg some use of the word "profile" still lingering in the latest editors draft that still need to be excised

Noah: schema 1.1 has added Schematron-like constraints that are closer to document structure. Schematron can represent things like business rules that do not mirror the document structure.

John: SML does not prevent other kinds of "definition" or "rule" documents from being part of an SML model.

Kumar: we have an implementation that you can validate a model with 1000 documents. Does take more time than just schema validation.

<MSM> [The submission had a section called 'XML Schema profile'; the current working drafts no longer have any such section]

John: regarding timeframe - have discussed all substantive issues; working now on getting these into the draft for Last Call.

Jonathan: Query WG should be interested in reviewing SML specs.

Andrew: comments from schema WG before LC is probably not possible.

John: likely in LC starting end of Feb and lasting at least a month

<Jonathan> ginny - i rather agree

W3C SML F2F Day 3 of 3

23 Jan 2008

Attendees

Contents

schemaComplete discussion for Bug http://www.w3.org/Bugs/Public/show_bug.cgi?id=5395/

Review of SML with Schema WG

Summary of Action Items