W3C home > Mailing lists > Public > www-tag@w3.org > November 2002

[Minutes] 18 Nov 2002 TAG ftf meeting

From: Ian B. Jacobs <ij@w3.org>
Date: Mon, 25 Nov 2002 17:30:08 -0500
Message-ID: <3DE2A470.7030902@w3.org>
To: www-tag@w3.org

Hello,

Minutes of the 18 Nov TAG face-to-face meeting in Boston
are available as HTML [1] and as text below.

   - Ian

[1] http://www.w3.org/2002/11/18-tag-summary

-- 
Ian Jacobs (ij@w3.org)   http://www.w3.org/People/Jacobs
Tel:                     +1 718 260-9447

==========================================================


W3C | TAG | Previous: 11 Nov teleconf | Next: 25 Nov
teleconf

      Minutes of 18 Nov 2002 TAG face-to-face meeting

Nearby: Teleconference details ? issues list ?
www-tag archive

1. Administrative

   1. Roll call: SW (Chair), TBL, TB, NW, DO, DC
      (Afternoon), CL, PC, IJ (scribe), Martin Duerst
      (Morning). Regrets: RF
   2. Did not yet accept 4 Nov minutes
   3. Accepted this agenda
   4. Next meeting: 25 Nov teleconf.
   5. No meetings: 23, 30 December.

1.1 Completed actions

    * Action SW, TB, DO: Send slides for AC discussion
      to TAG for review.
    * Publish HTML slides submitted by SW, TB, DO. TAG
      should comment on draft slide presentations on
      the TAG mailing list.
    * Submit three items to the Comm Team for the AC:
      TAG summary, SW's summary of XLink, Arch Doc.

1.2 February meeting

The TAG may reschedule its February meeting (both
days and location). Follow-up on the TAG list.

1.3 Presentation at W3C AC meeting

The TAG reviewed slides for its presentation at the
W3C Advisory Committee meeting (Nov 2003).

2. Technical

    * 2.1 URIEquivalance-15, IRIEverywhere-27
    * 2.2 Architecture Document
    * 2.3 RDDL, namespaceDocument-8
    * 2.4 Linking, xlinkScope-23
    * 2.5 Postponed

2.1 URIEquivalence-15, IRIEverywhere-27

   1. IRIEverywhere-27
        1. See reply from Paul Grosso asking the TAG to
	 address this issue quickly.
   2. Status of URIEquivalence-15. Relation to
      Character Model of the Web (chapter 4)? See text
      from TimBL on URI canonicalization and email from
      Martin in particular. See more comments from
      Martin.
        1. CL 2002/08/30: Ask Martin Duerst for
	 suggestions for good practice regarding URI
	 canonicalization issues, such as %7E v. &7e
	 and suggested use of lower case. At 16 Sep
	 meeting, CL reports pending; action to send
	 URI to message to TAG.

[Ian]

         SW: Time for a finding?
         TB: I think we are ready for a finding on
         matching semantics.
         CL: There are tough issues (e.g., proxy
         caches, etags) in different communities. I'd
         like to declare some issues out of scope so we
         can get quick agreement on smaller sets.
         TBL: I think we need to explain where IRIs fit
         in architecturally. My understanding is that
         there's a URI space and an IRI space, and a
         function that maps from one to the other. We
         can talk about equality among URIs, and we can
         talk about mappings from IRI -> URI. We need
         to point out that IRIs are a way to talk about
         URIs (like relative URIs map to absolute
         ones).
         TBL: We are not changing the URI space by
         talking about IRIs. There is not a function
         from URIs to IRIs.
         CL: Why would someone want to go from URIs to
         IRIs?
         MD: Sometimes you go to URIs to early.
         CL: But that's a repair issue.
         TBL: People *will* want to display URIs in IRI
         syntax.

[Chris]
         rrsagent, pointer?

[RRSAgent]
         See
         http://www.w3.org/2002/11/18-tagmem-irc#T14-44
         -48

[Ian]
         MD: If you define some URIs to be very
         strictly equivalent in all cases (including
         namespace matching) it makes it natural to
         define some IRIs to be equivalent to those
         URIs in all cases. Suppose U1 is equivalent to
         U2. It makes a lot of sense to say that I3 is
         always equivalent to them.
         TB: It's not, unless you fix the % escape case
         matching.

[Chris]
         I suspected they were not equivalent, but
         wanted to be clear

[Ian]
         TBL: If IRIs stand for URIs, then comparing
         IRIs doesn't make sense. You always talk about
         equivalence of URIs.: I advise you strongly to
         not redefine HTTP and MAILTO for IRIs...
         CL: To compare I1 and I2, you have to convert
         them and compare U1 and U2.
         [Discussion about escaping]
         TBL: We can give you a set of circumstances
         under which you can know that two URIs are the
         same. You know that if URIs are byte-by-byte
         equivalent, they are the same. We can add to
         that canonicalization of %7e and %7E. There
         are other things you can know from other specs
         that lead to equivalence. We should say "don't
         rely on that other stuff".

[Chris]
         CL was asking a question of TimBL not stating
         a position

[Ian]
         TBL: We suggest you use the same byte sequence
         for the same URI.: Basically, don't be clever.

         TB: Three ways to test URI equivalence:

        1. "Schema-specific canonicalization and
	 comparison" (e.g., HTTP URIs with
	 normalizing on "..").
        2. "RFC2396-based canonicalization". (e.g.,
	 maximum canonicalization, or lightweight
	 canonicalization such as canonicalization of
	 hex escapes)
        3. Compare ASCII character values.

         TB: All of these are potentially reasonable
         (though (c) is most questionable). The
         namespace Rec says "character by character"
         comparison, so there's some ambiguity. In any
         sane universe %7e and %7E are the same
         character. At the very least, we should
         document the multiple levels of comparision,
         and document the application contexts where it
         would be reasonable to do these things. And we
         strongly recommend that specs be entirely
         specific about what they require.
         TBL: NO! URI equivalence is not spec-specific.
         Equivalence is not a property of other specs.
         TB: You can't make the judgment calls go away.
         It's perfectly ok to not be sensitive to
         HTTP-specific issues when you're dealing with
         general URIs.
         SW: Different degrees of "equivalence". There
         are many equivalence relationships.
         TBL: String identity is the highest form of
         equivalence in our context. If you have string
         identity, it implies every other form of
         equivalence. Expanding circles are: ascii
         string THEN %7e <=> ~ THEN %7e <=> %7E THEN
         ".." and "." THEN HTTP/DNS THEN SMTP, etc. ...

[Chris]
         <foo bar="HelloWorld"/>
         <foo bar="Hello&#x57;orld"/>
         <foo bar="Hello&dubya;orld"/>

[Ian]
         TB: I think most appications would throw on
         the floor and xml doc with a namespace with
         W3.ORG in uppercase.
         CL: If the value of bar is "any URI", are
         those three the same? I think so, since they
         are the same strings after XML processing.

[Chris]
         after xml parsing al three are the safe and
         will match under option 3
         [Unanimous]

[Ian]
         IJ: Should the TAG say "Do up to and including
         level X?"
         TBL: Since practice varies we should not tell
         people they can rely on level 3.
         TB: But we could tell people for future
         practice to do a particular level.

[Chris]
         [Discussion on 2396 and whether it does
         foo/../bar canonicalization]
         TBL: relative uris require that /../ must be
         equivalent to /
         SW: want to see this explicit
         TB: RF is fixing this we believe

[timbl]
         TBL: RFC2396 dioes indeed not say that
         xxx/./yyy is equivalnet to xxx/yyy fopr any
         xxx and yyy. However, the only tenable
         situation is that they are equivalent. because
         we require that any URI can be relative-ized
         and absolute-ized back to its original. That
         is an (unspoken) axiom.
         When you relative-ize things and re-absolutize
         then, you cannot distinguih between the two,
         and so they HAVE to be equivalent. The URI
         spec should say that.
         Resolved: unanimously (except for IJ, who
         stepped out of the room).

[Ian]
         TB summarizing:

        1. we should talk about mapping from IRIs to
	 URIs and URI-based equivalence testing. Show
	 the Venn diagram of URI comparison levels.
        2. We may have agremeent about levels of Venn
	 diagram that specs should say.
        3. Explain drawbacks of doing different levels.

[Chris]
         Should also indicate strengths and drawbacks
         of different choices

[Ian]
         IJ: What does this mean for XML Namespaces?
         What story would you tell?
         TB: I think it would be helpful to tell the
         core group which level to use. Meanwhile, I
         don't think we can say when to use IRIs since
         they're not baked yet.
         IJ: But we can tell them how they fit in./
         CL: I think we can respond to initial query.
         We can tell people to plan on using IRIs when
         they're done.: We can tell people to design
         specs to use URIs, but to prepare for the
         introduction of IRIs. Tell them how to not
         paint themselves into a corner.
         NW: The XML Core WG is trying to decide what
         to put into the next version of the namespaces
         spec. They have a capsule summary of what IRIs
         are, to be replaced by a ref to the IRI spec
         later. I can't tell from the discussion here
         whether we are telling the core WG if it's ok
         to do what they're doing.
         Initial questions from Jonathan Marsh
         Action TB: Write a finding for
         URIEquivalence-15 on IRI relation to URI,
         different levels of equivalence.
         TBL: If a spec transitions from URIs to IRIs,
         new documents will break old software (which
         used to only handle URIs). Typically, what
         happens is that user agents evolve but authors
         SHOULD NOT use the new stuff for a while. We
         could tell people that it's ok for the
         software to accept IRIs but authors should not
         use them for a while.
         SW: I think we need a thought-out transition
         plan.
         CL: "anyURI" in the XML Schema spec is already
         IRI-ready.
         IJ: Specs should not handle this differently.
         Can we give them text for their documents?
         MD: Some specs may have to do this
         differently.
         NW: People use "stringcmp" to compare
         namespace strings.
         SW: What should our response to Jonathan Marsh
         be?
         TB proposal:

        1. We view IRI activity with favor.
        2. Software should prepare for IRIs
        3. IRI spec not done, practices such as XML 1.0
	 sys id seem to be reasonable, but they need
	 to figure out how to bring themselves into
	 sync with IRIs when they become available.

         MD: I agree that developers need to think
         about the transition issue.
         TBL: We can't tell people what's best to do
         for their applications (since so many
         variables, policies, etc.). We can describe
         the framework.
         CL: We need to explain the risk of not moving
         to IRIs: proprietary approaches if IRIs not
         adopted.
         PC: XML Query has an "anyURI-equal" funtion.
         Described as being on a
         "codepoint-by-codepoint" basis.
         PC: If you are looking for behavior of anyURI,
         don't look in schema, look in query. See
         XQuery definition of an op:anyURI-equal
         function See XQuery definition of an
         op-resolve-uri function.

[timbl]
         TB, please point to the layer of the onion
         which corresponds to schema:anyURI.Equal()

[Ian]
         PC: See Functions and Operators.
         NW: For namespaces 1.1, the
         backwards-compatibility issue already exists.
         It's safe to say "they can be IRIs"; breaking
         software is probably not a huge issue yet.
         TBL proposal:

        1. Specs (e.g., an XML application) that use
	 Unicode should call out IRIs.
        2. Refer to the upcoming IRI spec, which we
	 hope will stabilize soon.
        3. Warn people that authors should stick to
	 URIs during a transition period, that will
	 vary according to their transition period.

         CL: Namespace 1.1 should say that a particular
         usage is being brought up to date with other
         usage.
         TBL: We need to write down the axioms: if you
         take a URI, make it relative w.r.t. a base
         URI, then make it absolute w.r.t. the same
         base URI, you get the same starting URI...
         CL: I think the finding should explain pros
         and cons as we said above, what people with
         old specs should do, and what people with new
         specs should do.
         Action MD: Write up text about
         IRIEverywhere-27 for spec writers to include
         in their spec.
         Action CL: Write up finding for
         IRIEverywhere-27 (from TB and TBL, a/b/c), to
         include MD's text.

2.2 Architecture Document

See also: findings.
   1. Findings in progress:
        1. deepLinking-25
	   1. TB 2002/09/09: Revise "Deep Linking" in
	      light of 9 Sep minutes. Status of
	      finding?
   2. Arch Doc
        1. Continued action CL 2002/09/25: Redraft
	 section 3, incorporating CL's existing text
	 and TB's structural proposal (see minutes of
	 25 Sep ftf meeting on formats).
        2. Completed action NW 2002/09/25: Write some
	 text for a section on namespaces (docs at
	 namespace URIs, use of RDDL-like thing).
	 Done
        3. Continued action DC 2002/11/04: Review
	 "Meaning" to see if there's any part of
	 self-describing Web for the arch doc.

See 15 November 2002 Architecture Document. See some
comments from Tim Bray.

[Ian]
         TB: In the principles, the distinction between
         "constraints", "practices", and "principles"
         still needs work. Perhaps we can move
         simply to "practices" and "principles" - it's
         really unclear that "Use URIs" is really
         different in its nature from a bunch of things
         labeled "practices".
         [Some disagreement from CL, TBL on this]
         TB: The principles in 2.2.4 and 2.2.5 are
         really the same principle. The explanatory
         text in 2.2.5 is just a rehash of the Moby
         Dick example.
         IJ: I can explain why this was done: I have
         been trying to distinguish case of
         "representations that vary inconsistently" and
         "meaning of use" discussions.
         TBL: Previous draft for me was too vague. Too
         general to say that "ambiguity is a bad
         thing". To say that a mailto URI identifies a
         book is wrong. The RFC says different. To
         couch that as ambiguity is wrong. One spec
         owns the definition.
         TB: In the real world, the marketing dept and
         the IT dept may use URIs internally
         differently.
         SW: What about namespace names. We are
         expecting to put documents at the end of the
         namespace URI. In RDF, how do I make
         assertions about the document and assertions
         about the namespace.
         TBL: You have to be able to deal with the
         different levels.
         TB summarizing:

        1. We agree that ambiguity is bad.
        2. When dealing with these things, you follow
	 specs (e.g., don't use mailto URIs to
	 identify things other than mailboxes).

         SW: The generic statement is "follows specs".
         IJ: Please confirm that there is some
         authoritative meaning; I can only understand
         ambiguity w.r.t. some reference.
         TB: The arch of the web talks about resources
         and representations.: 2.2.4 as written is ok.
         It implies inconsistency in identity of
         resource. The other issue is that using the
         web in a way that goes against specs is wrong.
         IJ: You can use mailto uris to talk about
         mailboxes. I presume you can also use http
         URIs.
         TBL: Not to refer mailboxes as defined by RFC
         2368
         TB: Ambiguity is bad, on server or client.
         Don't fly in the face of specs.
         TBL: We need to also highlight *design
         choices* in the architecture document.
         TB: Two other suggestions: Arguing about the
         range of URIs will not be useful right now. I
         have some proposals for sections 3 and 4.
         DO: I think we haven't resolved yet what are
         the components of this document (constraints,
         principles, etc.).

[Chris]
         3.4 third list item "Allow for Web-wide
         linking, not just internal document linking."
         I would like to discuss that as low hanging
         fruit.

[Ian]
         Discussion ofproposals from Tim Bray.
         IJ: I am not currently working on a "model"
         that fits together constraints, principles,
         etc. Too hard to rip up the document at this
         time.
         PC: I think a glossary of those terms would be
         useful.
         IJ: The terms I had listed were: constraint,
         principle, design choice, good practice,
         required property.
         TB: Sections 3 and 4 have been languishing. We
         should just start putting in nuggets and then
         fill in with language around them.
         Proposed CP1 "When designing a data format to
         be used in representing Web Resources, the use
         of XML should be considered carefully. - some
         issues concerning XML pros and cons, - refer
         to IETF 'Guidelines for the Use of XML within
         IETF Protocols'."
         CL: That document does describe pros and cons.
         IJ: Add XML Accessibility Guidelines for use
         of XML in protocols doesn't talk about
         accessibility.
         Action IJ: Send notes to TAG with comments on
         using xml.
         Resolved: Add CP1 to spec.
         Proposed CP3 "When specifying the use of URIs,
         designers SHOULD NOT constrain the
         use of URI schemes."
         TBL: In some constrained applications, you may
         want to e.g., constrain to some class of URNs.
         CL: What about something like "You must
         support the HTTP protocol on this element, and
         may support others".
         TBL: Yes, that's ok.

[Chris]
         Except for phones that haver no http stack?

[Ian]
         Resolved: Add CP2 to spec. Provide a
         counter-example.
         Proposed CP3: "When using XML, designers
         SHOULD NOT introduce syntax constraints beyond
         those involved in the definition of XML."
         CL: Different specs impose additional
         syntactic constraints (e.g., namespaces).
         TB: E.g., what SOAP did about not having an
         internal subset is wrong. SOAP imposes a
         severe cost (you can't use an off-the-shelf
         XML parser). You can't enforce the SOAP
         constraints by using off-the-shelf products.
         DO: I disagree with the principle.
         TB: There's a big difference between
         profiling, and saying "you can use single
         quotes only but not double quotes".
         Resolved: CP3 rejected as proposed.
         Proposed CP4: "XML-based languages MUST be
         given a namespace name and the elements of the
         language MUST be placed in that namespace.
         Designers SHOULD make available a
         representation of the namespace which is
         human-readable and SHOULD make available a
         representation which is a machine-readable
         directory of resources which are related to
         that namesapce."
         Amendment: "XML-based languages for widespread
         common use MUST be given a namespace name and
         the elements of the language MUST be placed in
         that namespace."
         PC: What about data types and functions?
         TB: There's room for argument about things
         other than elements; there are other design
         choices.
         PC: An XML-based language might have other
         components. Does the arch doc not need to make
         statements about those other components?
         TB: XML namespaces only refers to elements and
         attributes, not other types of things.
         Amendment: "XML-based languages for widespread
         common use MUST be given an XML namespace name
         and the elements of the language MUST be
         placed in that namespace."
         TBL: I agree with TB's more specialized
         statement, but I think that important
         resources such as those functions and
         operators need to be identifiable by URI.
         [Discussion that TAG has not yet agreed on all
         of second sentence of CP4.]
         Resolved: Accept "XML-based languages for
         widespread common use MUST be given an XML
         namespace name and the elements of the
         language MUST be placed in that namespace."
         SW: Can it draw on elements from other
         languages?
         TB: If you important elements from another
         language, that's fine.
         TBL Proposed: "If you are designing and XML
         language, in which the required functionality
         is available from elements in another
         namespace, there is benefit from the reuse
         those elements."
         CL: There's a problem of content types.
         General agreement for a statement encouraging
         the reuse of previously defined elements where
         appropriate.
         CL: I had wanted style properties to be in a
         namespace....
         TB: CL, I suggest you write down a principle.
         Proposed CP5: " For languages whose contents
         are intended for rendering to humans, the
         repertoire of formatting semantics SHOULD be
         consistent across the universe of W3C
         recomnmendations."
         Resolved: Accept CP5, with examples (e.g.,
         style sheets). Link to relevant finding (and
         similarly for other of these proposals).
         Proposed CP6: "When designing a language that
         includes linking or hypertext functionality,
         designers SHOULD design that functionality so
         it supports Web-side rather than merely local
         linking."
         CL: About CP6 - in SVG you can point from a
         fill to a gradient. We could have made this an
         idref, but we made it a link so you can reuse
         gradients.
         TB: Another way to say this is "Don't use
         IDREF"....
         DO: In SOAP, they use ID and IDREF.
         NW: Does anybody here disagree with CP6?
         Resolved: Accept CP6.
         Proposed CP7:"Designers of languages which
         will be used in resource representations MUST
         arrange for the registration of an Internet
         Media
         Type for that language, and SHOULD consider
         the recommendations of RFC3023 in carrying out
         that registration. This registration MUST
         include a specification of the handling of
         fragment identifiers for resource
         representations in the language being
         designed."
         [TB notes that we have a finding on this.]
         CL: Note that "+xml" does not define fragment
         semantics.
         TB: I don't think we should rely on default
         semantics. Be explicit if you expect to change
         semantics later.
         DO: Indicate that there is no default fragment
         identifier semantics for XML.
         TBL: See RFC3023: "As of today, no established
         specifications define identifiers for XML
         media types. However, a working draft
         published by W3C, namely "XML Pointer Language
         (XPointer)", attempts to define fragment
         identifiers for text/xml and application/xml.
         The current specification for XPointer is
         available athttp://www.w3.org/TR/xptr."
         CL: My worry is not that the spec doesn't say
         something....
         DO: If XPointer goes to Rec, will RFC3023 be
         revised?
         TB: Even if XPointer goes to Rec, that doesn't
         help much. You don't know what ID elements are
         unless you have a DTD.... For CP7, the point
         is "read RFC3023 and think about it."
         Action CL: Incorporate resolutions above into
         a proposal for chapter 3. [Scribe presumes
         this supersedes previous action from
         2002/09/25]

LUNCH

DC arrives.

[RRSAgent]
         See
         http://www.w3.org/2002/11/18-tagmem-irc#T18-32
         -01

[Ian]
         Resolved: Accept CP7 with language about
         warning about no default meaning for frag ids.
         [See below for amendment.]
         DC: WebOnt WG is currently wondering what
         media type to use. Owl? I conclude that
         application/rdf+xml is the right thing
         TB: You should arrange for the registration
         "unless one existst that you can use."
         [Question: "What MIME type should I use?"]
         TBL: If you know that the XML app is going to
         dispatch on the root element, you can use
         text/xml with impunity. RF pointed out that
         for the infrastructure, it's useful for, e.g.,
         SVG, to have its own mime type. We agreed that
         it's better to dispatch on the mime type.

[Chris]
         ... where the best choice is an existing media
         type

[Ian]
         DC: I think that CP7 will lead to more
         questions from WGs.
         TB: I think it's useful to say that languages
         at W3C should have mime types unless there's
         really a good reason not to.l
         Resolved: In CP7, say "SHOULD arrange" instead
         of "MUST arrange".
         Proposed CP8: "Designers of languages which
         are to be interchanged on the Web MUST include
         a discussion of error-handling, with specific
         recommendations on the correct behavior upon
         detection of certain classes of errors. -
         example classes: XML well-formedness vs.
         semantic brokenness (eg SVG circle with
         negative radius)"
         CL: SVG spec can't revise error handling of
         XML spec, though.
         DC: What about general principle about being
         liberal in what you accept and conservative in
         what you produce?
         CL: No.
         DC: Then I object to this point.
         TB: RFC3023 includes some discussion of this.

[DanC]
         The word "liberal" does not occur in RFC3023.

[Ian]
         DC and CL agree that the liberal/conservative
         Internet principle should be mentioned because
         IT IS NOT universal.

[Chris]
         "Forbid working around xml well-formedness"

[Ian]
         IJ: I find CP8 too broad.
         DC: No, there's something more basic: Don't
         silently throw away error messages.
         TBL: It only makes sense to define processing
         where the protocol gives you some guarantee
         through the processing.
         CL: SVG has error-handling.
         TB: I think W3C specs shouldn't advance if
         they don't discuss error-handling.
         DC: To me, the core principle is about
         evolvability or scalability.
         Resolved: CP8 rejected as proposed.

2.3 RDDL, namespaceDocument-8

See 8 Nov 2002 version of RDDL from TB and Jonathan
         Borden and issue namespaceDocument-8.

[Ian]

         TB: RDDL goes back several years. Some
         principles: (1) be human-readable (2) be able
         to find style sheets and schemas and other
         stuff. Need metadata: (1) purpose and (2)
         nature. I should be able to say "Get me an xml
         schema" and use the metadata in the RDDL
         document to find one (or one from several,
         according to additional purpose or nature
         constraints). Natures and purposes are URIs.
         nature-> xlink:role, purpose->xlink:arcrole
         Some disputes about that choice. I created a
         RDDL in RDF draft. But some people pointed out
         that this was done in a way that wasn't
         kosher.

[Chris]
         this is an ongoing discussion n xml dev
         what should we do?
         already committed to write a note, using
         rdddl, could have the xlink encoding or the
         rdf
         in the latter case, lots more discussion
         needed
         or use another RDDL namespace
         first draft should do all three and discuss
         dc: use cases where this would help?
         tb: ms office 11 stores everything as xml, it
         is al well formed and uses namespaces heavily
         also, stop using urns, use http instead and
         point to persistent urls and point to schemas
         RDDL would be handy for this
         point to stylesheets, .net etc etc
         dc: why not point to these from the schema?
         cl: same reason as not pointing to the schema
         from the .net everyone wants to be top
         so a neutral, easy to parste format to point
         to these
         dc: why mix them up
         tb: because the one url serves this
         dc, tbl: use coneg
         tbl: or put in html, advantage is mixomg
         metadata and html
         tbl; many sites do not do coneg anyway

[TBray]
         also many people who publish to webservers
         don't have control of the conneg settings on
         the server

[PaulC]
         Latest Q&A on MS Office 11 XML support.

[Chris]
         cl: cause for concern of using a
         multi-namespace xml design in a system (html
         browser) that is not xml
         dc: html wg should be part of the design
         skw: who is the authorship of this
         tb: jb and i to publish as a w3c note
         do: wsdl havea requireent to identify the
         elements that are semantically interesting. If
         a document is replicated, it no longer has a
         unique uri. Namespace name is consistent
         across these representations.

[DanC]
         reviewing minutes from Sep ftf, we didn't
         decide anything about RDDL; we just actioned
         TB to propose something.

[Chris]
         do: Want to use ns name as base uri for these
         constructs, eg port type. How to write a uri
         reference that identifies the port type?
         without ids. Don't want to generate ids for
         all elements. NS name of the service, append
         the construct
         [do draws on whiteboard]
         piza delivery web service
         wsdl specifies interface info (astract) and
         physical deployment
         two pizza shops have to publish at different
         urls
         do: so .... multiple users of one wsdl file,
         all clients have the same instance
         no single master url for this document
         this is the 'locally stored copy' issue
         only common thing is the namespace url of the
         service
         they did GET copies
         just, not every time they want a new pizza

[DanC]
         er... he said the didn't. or at least: I heard
         that.

[Chris]
         do: so they want a wsdl-specific xpointer
         scheme - but that depends on the media type,
         so we cant use this inside a xhtml document in
         there

[DanC]
         (it's sooo frustrating paying the cost of not
         just using RDF)

[Chris]
         so need to dereference through the rddl
         document to their scheme
         dc: ok I see the problem
         tbl: not a problem unless you use html. RDF
         uses barenames not xpointer schemes, but also
         appends onto namespace names, but keeps the
         same ns name for all the multiple copies

[Ian]
         TBL: There's a solution - say that for HTML
         elements, # refers to element, but for WSDL,
         refers to a real-life thing.

[Chris]
         skw .wsdl means you could have a separate
         media type
         dc: but rddl in the middle would break that
         connection

[Ian]
         DO: If we go down RDDL path, we need to define
         frag id semantics.

[Chris]
         tb: he wants t point into his rddl doc using
         his xpointer scheme
         cl; ok, but that means he can only point to
         wsdl documents
         [dc writes on whiteboard]
         dc: schema for html block elements. Schema for
         p, p extends block, etc.
         urlofdoc#p
         no, because urlofdoc#style
         is that the style elemnt or the style
         attribute?
         xptr can point into this
         RDF answer is to ensure its a directed
         labelled graph
         then, turn the crank and get the syntax
         pc: does not solve the anonymous types issue
         nw: nun ensures that all the anonymous types
         have distinct uris
         dc: so now we see the cost of wsdl not doing
         this
         do: referring to wsdl issue 120 about unique
         adressability was raised by SW folks. WSDL has
         requirements above the requirements of SW.

[Roy]
         I'm at ApacheCon in Las Vegas. No phone, but
         will hang out on IRC if there are any
         questions. [No opinion on RDDL]

[PaulC]
         XML Schema Nun proposal (W3C member only)

[Ian]
         Specific example from Jonathan Borden:
         <rddl:resource ID="XSD">
         <rddl:title>XML Schema</rddl:title>
         <rddl:nature
         resource="http://www.w3.org/2001/XMLSchema"/>
         <rddl:purpose
         resource="http://www.rddl.org/purposes#schema-
         validation"/>
         <rddl:related
         resource="http://example.org/L.xsd"/>
         <rddl:prose>
         <p>An XML Schema for the L language .</p>
         </rddl:prose>
         </rddl:resource>

         TB: I want to be able to reach into a RDDL
         document and find "purpose: validation". I
         want to be able to reach into a RDDL document
         and find "purpose of validation".

         DC: In RDF you would say
         "purposes:validation". Nature and purpose are
         not symmetric.
         TBL: The community mismodeled nature and
         purpose.
         TB: Whenever you put in something more
         complicated than
         http://lists.xml.org/archives/xml-dev/200211/m
         sg00719.html, people get unhappy.

[DanC]
         <rddl:resource ID="XSD">
         <rddl:title>XML Schema</rddl:title>
         <rddl:nature
         resource="http://www.w3.org/2001/XMLSchema"/>
         <p:schema-validation
         resource="http://example.org/L.xsd"/>
         <rddl:prose parseType="Literal">
         <p>An XML Schema for the L language .</p>
         </rddl:prose>
         </rddl:resource>

[Ian]
         PC: Schema WG has rejected using namespace
         name for locating frag ids. Sounds like
         multiple WGs tackling same problem with
         different solutions.

[PaulC]
         XML Schema Designator of Schema (Member only)

[DanC]
         p is http://www.rddl.org/purposes#

[Ian]
         TB: Jonathan suggested what DC suggested.
         DC: Problem with the RDDL approach is that I
         can't cut and paste directory entries (or
         merge them).

[DanC]
         The "implicit" design contradicts the "anyone
         can say anything about anything" principle of
         RDF

[Ian]
         CL: But we *are* talking about namespace docs,
         so the namespace URI is implicit.
         DC: But the namespace doc is a common place to
         look for it. It doesn't mean that you
         shouldn't be able to copy it and preserve the
         meaning. You could change "id" to "about' and
         put the namespace there.
         PC: Perhaps we need some more time to discuss
         this.
         TB: Note that
         http://lists.xml.org/archives/xml-dev/200211/m
         sg00719.html has received a lot of support. We
         should not make more complicated than we need.
         Dan's representation is more accurate.
         NW: If you are using predefined purposes, you
         could avoid extra namespace.
         TB: DC's version makes purpose less obvious.
         RDDL was sold to the world as having nature
         and purpose.
         CL: I propose that the first draft show
         different possible syntaxes, with their pros
         and cons.
         PC: I think that the work done on this by
         Schema (Member-only) is material to this
         discussion.
         TB: For the record, I promoted RDDL as a way
         to do a 2-field lookup. I'm not going to be in
         favor of requiring any bending over backwards
         to accomplish this effect.
         [DC proposal]
         Suppose L is a namespace about lunch.
         Namespace URI: http://example.com/L. There are
         two related resources: L.xsd and L.css.
         TB: L.xsd is a nature (xml schema) and a
         purpose (for validation).
         [DC does circles and arrows graph]
         TB: Add a CSS style sheet as well. Purpose is
         "Onscreen presentation". There's a reserved
         word for that in RDDL... The nature of the
         style sheet includes IANA stuff, and
         more...(Arguably there's a better way to do
         this).
         DC: Two resources are related by "purpose".
         TB: Sample application - build a menu of
         available actions in a RDDL document. I'll buy
         DC's graph. However, using RDF syntax I would
         need an RDF engine.
         DC: So we need to pick one syntax.
         TBL: We could propose a canonical xml
         serialization of RDF.
         CL: application/rddl+rdf?
         TB: That's a hard problem, too..
         CL: We have two contradictory principles for
         W3C specs - you need to validate v. mixing
         multiple namespaces.
         TB: I personally think RDDL is important. I
         think we need to converge quickly on a
         suggested right way to do this.
         Action TB: Solicit proposals for what a
         namespace doc should look like for the
         particular case of a namespace document that
         refers to a DTD and a style sheet. TB will
         collate the responses. TB's solicitation
         should ask submitters for their own pros and
         cons.
         Action NW: Take a stab at indicating pros and
         cons for the various designs.

2.4 Linking, xlinkScope-23

Seeissue xlinkScope-23.

[Ian]
         PC: What's our next step?

         TBL: I think there's a clear need for a common
         anchor element in the XLink namespace.
         NW: Content model problem....
         CL: OBJECT has three URIs and two bases. Can't
         use xbase.
         TB: Maybe right answer is the summit at the
         tech plenary
         PC: That may not happen.
         TB: This is pressing. People want multiend
         links and are hacking horribly to do this.
         NW: When XLink was done, it was made a little
         too general.
         TB: See my concrete proposal, with support
         from Ann Navarro
         [Scribe stops minuting since discussion and
         scribe a bit unfocused]
         PC Summarizing some options for moving
         forward:

        1. We listen to the AC tomorrow and Weds
        2. Work on a charter, and send to AC for
	 discussion.
        3. Wait for special meeting in March (if it
	 happens). But several people think this is
	 too late.

         NW: I think we should do this sooner rather
         than later. We should convene interested
         parties before Mar 2003.
         DC: We could invite people to our Feb ftf
         meeting to discuss this.

[Chris]
         html, smil, xforms, svg
         Action SW: Create such a special-interest
         telcon

         The following people committed to attending
         such as meeting: TB, CL, NW.

         DC: people who are the right hot people should
         be there. Docbook folks, too. Client builders
         (e.g., Tantek Celik and others). Ask for
         attendance by way of the HMTL Coordination
         Group and the XML Coordination Group.

2.5 Postponed

   1. contentPresentation-26
        1. Action CL 2002/09/24: Draft text on the
	 principle of separation of content and
	 presentation for the Arch Doc.
   2. rdfmsQnameUriMapping-6
        1. The Schema WG is making progress; they will
	 get back to us when they're done. See XML
	 Schema thread on this topic.
   3. uriMediaType-9:
         + Action DC 2002/08/30: Write a draft Internet
	 Draft based on this finding (Deadline 11
	 Nov). This action probably subsumes the
	 action on TBL to get a reply from the IETF
	 on the TAG finding.
   4. Status of discussions with WSA WG about
      SOAP/WSDL/GET/Query strings?
         + DO 2002/06/24: Contact WSDL WG about this
	 issue (bindings, query strings and schemas)
	 to ensure that it's on their radar. See
	 discussions from 9 Sep TAG teleconf.
    ________________________________________________


   Ian Jacobs, for TimBL
   Last modified: $Date: 2002/11/19 16:12:39 $
Received on Monday, 25 November 2002 17:30:12 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Thursday, 26 April 2012 12:47:13 GMT