W3C

- DRAFT -

Technical Architecture Group F2F

07 Jun 2011

Agenda

See also: IRC log

Attendees

Present
Noah Mendelsohn, Jeni Tennison, Henry Thompson, Daniel Appelquist, Yves Lafon, Ashok Malhotra, Larry Masinter, Jonathan Rees, Tim Berners-Lee
Regrets
Peter Linss
Chair
Noah Mendelsohn
Scribes
Yves Lafon, Daniel Appelquist

Contents


Fragment ID Semantics

http://www.w3.org/2001/tag/2011/06/06-agenda.html#mimefrag

Jeni describes the goal of the session

rfc3986 describe fragments and how they relates to meia types, AWWW talks also about fragment and conneg

3023bis draft describe that fragments MUST be interpreted per XPointer spec

- hashbangs and conflict with HTML/XHTML media type definitions

Larry: HTML5 contains new definition for text/html type

Jeni: media fragment define fragment syntax, but SVG can't use it because of rfc3023bis rules

<masinter> suggest looking at some other media type definitions with fragment identifiers for completeness and additional constraints

In SemWeb fragments can describe unconstrained things

which is an issue for the application/rdf+xml media type

<masinter> think there were proposals for letting fragment identifiers to depend on the scheme as well as the media type, especially for streaming protocols

RDFa core is a way to add RDF in XML documents, leading to situations where fragments won't necessarily identify parts of the document

jar: we should need coordination with 3023bis for the RDFa core case.
... when you have RDFa that identify endpoints that may be fragments of the XML document

Jeni: new media type for other RDF serialization are silent on fragments

<noah> NM: Does anyone have a formal action to follow up on "agreement" with 3023bis editors regarding changes for fragid and generic processing?

<noah> HT: I informally track 3023bis progress (Henry later clarified in private communication that there is no formal action for his tracking)

<noah> JAR: Let's do this at the end of this session.

<noah> NM: OK, as long as we don't forget to get a formal action.

Jeni: first point is about SVG. Per media type definition, it has to use XPointer for its fragment identifier, however mediafragments describe ways of identifying part of an image => clash

Noah: we have two level of genericity there, the +xml and in media fragment it is about the image/ part

for application/svg+xml it is covered by 3023bis only

<masinter> i think the media fragment draft is inconsistent with other specifications

http://www.w3.org/TR/SVG/mimereg.html

<jar> http://www.w3.org/TR/2011/WD-media-frags-20110317/#standardisation-URI-fragments

http://www.iana.org/assignments/media-types/image/index.html links to image/svg+xml

<ht> http://www.w3.org/2006/02/son-of-3023/latest.html#naming

ht: from 3023bis image/svg+xml is covered by the +xml

dka: as an user, you will use the kind of fragment you need to idenfify part of XML or part of the image

ht: it is entirely fine to have two layers and be able to address each layer that are only disambiguated by use.

ht: good practice is "if you use XML representation, it make sense to use a +xml media type and reference bits of XML as XML" and have a syntax to do other non-syntactically overlapping identification per media type

noah: there is an issue that people want to say 1/ it's an image, 2/ it's XML 3, it's SVG

ht: issue with application/rdf+xml is that there is syntax overlapping between the RDF and the XML cases

<masinter> things are inconsistent (a) what are the possibilities and (b) what specs would have to change

timbl: SVG also have their viewport syntax

Jeni: it is expressed as valid Xpointer syntax

<masinter> discussion about how SVG uses an xpointer extension

<noah> YL: What if I'm an Xinclude processor processing an image/svg+xml

<masinter> could someone explain the xinclude/fragid connection?

<jar> RDF/XML gives a meaning to id="foo" ...

<ht> http://www.w3.org/2006/02/son-of-3023/latest.html#naming

ht: the 3023bis spec says "you should a +xml media type and the consequences are those..." for XML content

<masinter> thinking about the +json convention as another case

<masinter> ((yves asking about whether generic processors come across SVG fragment which isn't valid XPointer))

timbl: is there a way, if you have an id to make it explicitely clear that it's a Xpointer ID and not 'foo'

ht: yes, but nobody does that

timbl: so can you enforce non-overlapping by forbidding the use of bare names for XML content when you mean xpointer id?

ht: there are lots of pointers, so doing so would be very difficult

<masinter> are there any other cases of inconsistencies between the generic URI/MIME etc. approaches and the specific case

ht: people want sometimes to point to XML parts, sometime to media-type specific fragments (like SVG viewport)

<masinter> text/* vs. text/html and charset, EOL, line length limits

<jar> conneg between python and common lisp?

jeni: we are in agreement that there is an issue with the wording of 3023bis that makes it stricter than needed the fragment processing, and it should talk about non overlapping syntax instead

larry: issue is that there are documents who are saying confilcting things and they all need to be fixed

another issue is text/* and defaut charset

jeni: what can we learn from other conflicts?

<jar> RDF id="foo" is part of the reification feature, and the RDF WG charter says "Deprecate some RDF features (e.g., reification...)" http://www.w3.org/2011/01/rdf-wg-charter

larry: "stuff happens", we can either document inconsitencies, try to fix the media type definitions...

noah: one think we can do is alert for future media type definition

<jar> well... maybe ... id= is used in 2 ways ... long story

for the RDF case we can document the breakage

timbl: there is real damage that needs to be fixed, when you have a mixture of XML and RDFa and anchors

<masinter> i want to make a case for not trying to fix fragment identifiers

<masinter> because fragment identifiers have such difficulty anyway, e.g., with non-ascii characters and IRI of fragments

timbl: it wold be tempting for an browser implementation or an RDF implementation to pre-suppose things about fragments based on their understanding of them

<masinter> encourage people not to use URIs with fragments to identify components, and instead use some other syntax, e.g., (content + xpointer) or (content + time syntax)

jar: three cases: 1. fragids can have different kinds of meaning in disjoint syntactic spaces, 2. fragids can have different meanings for different ids, 3. single fragid can have multiple meanings depending on processor

jar: Manu's example is #3

ht: SVG would be fine with #2, SVGView is not overlapping and is application specific syntax

there is a use case to have fragments that can be understood in two different ways _on purpose_

jeni: when he has an URI in #foo, what is meant?

jar: it depends on the processor

larry: would it be ok to warn people against using fragment identifier ?

<jar> all sorts of problems

<noah> NM: +1 to LM, with a big caveat. 10 years ago, he would have been exactly right. The RDF train, and many others, have long since left the station.

larry: issue with fragment microformats, i18n issues etc...

<jar> pain from having to shove multiple semantics into narrow pipe of #

jeni: we have to live with fragments

timbl: is the proposal to have another attribute in html or xml to have a kind of fragment? How can you bookmark that?

<jar> masinter: why would a bookmark have to be only a raw URI?

timbl: awww says that you should identify things using URIs

ashok: one solution would be to say "don't use generic media types definitions"

noah: our first resolution was to remove the genericity part, but the feedback from the XML community is that the genericity part was the important thing

<masinter> you SHOULD identify resources with URIs, not clear it says you SHOULD identify fragments with URIs too

http://www.w3.org/2008/WebVideo/Fragments/WD-media-fragments-spec/#naming-space

<masinter> new URI scheme media-fragment: which uses some other character than #

media fragment mandates the use of '=' which is not an allowed character in #1, so we can tell by inspection the kind of processing that should occur

larry: SVG can have script, so fragments can be identifying script state

timbl: RDFa can be embedded in SVG as well

<jar> timbl: Anchors, RDFa, and scripts all use the ncname syntactic space

<jar> timbl: How about if script parameters are always distinguished so they're in a disjoint syntactic space

<masinter> compound document is a separate issue

<jar> yves: We're getting into compound document issue - multiple kinds of content in one document - and that's a different issue

<jar> timbl: The compound document situation is the core problem

<masinter> what about IRI issues of fragments, ID with %xx in it cannot be referenced?

ht: the RDFa case and script case is generic, while SVG case is about what the media type definition says

noah: we spent time to talk about compound documents, here it seems that we are diving on identification of those documents

jeni: what are the next steps then?

<noah> ACTION-509?

<trackbot> ACTION-509 -- Jonathan Rees to communicate with RDFa WG regarding documenting the fragid / media type issue -- due 2011-06-15 -- OPEN

<trackbot> http://www.w3.org/2001/tag/group/track/actions/509

<noah> ACTION-543?

<trackbot> ACTION-543 -- Jeni Tennison to propose addition to MIME/Web draft to discuss sem-web use of fragids not grounded in media type -- due 2011-06-15 -- OPEN

<trackbot> http://www.w3.org/2001/tag/group/track/actions/543

timbl: should media fragment use Xpointer syntax?

to reduce the syntax spaces

<masinter> i think this belongs in a URI/IRI spec update

ht: wrt 3023bis, it is entirely possible that we would need something new in the future, so there are nothing we agreed on that we should require from the editors

<noah> ACTION Henry to track fragid issues in 3023bis, report to TAG and/or communicate with 3023bis editors as appropriate

<trackbot> Created ACTION-564 - Track fragid issues in 3023bis, report to TAG and/or communicate with 3023bis editors as appropriate [on Henry Thompson - due 2011-06-14].

Jeni: wrt 543 my text is about describing the issues, not about solving it

<noah> ACTION-543?

<trackbot> ACTION-543 -- Jeni Tennison to propose addition to MIME/Web draft to discuss sem-web use of fragids not grounded in media type -- due 2011-06-15 -- OPEN

<trackbot> http://www.w3.org/2001/tag/group/track/actions/543

<jar> Group decision to table discussion at point where Larry started talking about how MIME & Web doc is progressing.

<jar> Will resume later in this meeting.

IAB - TAG Meeting

<noah> We are dialed, but on break. Will be ready for start at the half hour

<noah> Entry from TAG agenda for this call: http://www.w3.org/2001/tag/2011/06/06-agenda#iab

<noah> TAG agenda points to this working agenda for the call: http://lists.w3.org/Archives/Public/www-tag/2011Jun/0012.html

(round of introduction)

<noah> We should be logging IAB attendee names, I think

http://www.iab.org/about/members.html

<noah> Olaf Kolkman

<noah> Jon Peterson

<noah> Alissa Cooper

<noah> Spencer Dawkins

<Bernard> Bernard Aboba

<ht> Dow Street

<noah> Russ Housley

<noah> Philippe Le Hégaret

<ht> Cindy Morgan

<ht> Regrets from Hannes Tschofenig

http://www.w3.org/2001/tag/2011/06/06-agenda.html#iab

<noah> http://www.w3.org/2001/tag/

<noah> TAG's public home page: http://www.w3.org/2001/tag/

ht: I am asking the two chairs an elevator pitch to decribe what each group is doing

<noah> Charter: http://www.w3.org/2004/10/27-tag-charter.html

<Bernard> http://tools.ietf.org/html/rfc2850

<Bernard> IAB Charter

<noah> The mission of the TAG is stewardship of the Web architecture. There are three aspects to this mission:

<noah> to document and build consensus around principles of Web architecture and to interpret and clarify these principles when necessary;

<noah> to resolve issues involving general Web architecture brought to the TAG;

<noah> to help coordinate cross-technology architecture developments inside and outside W3C.

<noah> Architecture of the World Wide Web: http://www.w3.org/TR/webarch/

<Bernard> From Section 2 of RFC 2850:

<Bernard> The IAB is chartered both as a committee of the IETF and as an

<Bernard> advisory body of the Internet Society. Its responsibilities include:

<Bernard> (a) IESG Appointment

<Bernard> (b) Architectural Oversight

<Bernard> (c) Standards Process Oversight and Appeal

<Bernard> (d) RFC Series and IANA

<Bernard> (e) ISOC Liaison

<Bernard> (f) External Liaison

<noah> http://www.w3.org/2001/tag/findings

<masinter> http://tools.ietf.org/html/rfc2850#section-2

<noah> Public TAG mailing list: http://lists.w3.org/Archives/Public/www-tag/

Bernard: IAB also has administrative duties, including IESG appointement, see section 2 above

series of document published on different aspects of the Internet

<Bernard> Yes

first topic: Versioning: IAB document on extensibility vs. TAG notes on versioning

<masinter> http://tools.ietf.org/html/draft-iab-extension-recs

<masinter> sent to TAG http://lists.w3.org/Archives/Public/www-tag/2011Feb/0141.html but not on our agenda

Bernard: it is an attempt to document ways of extending protocols

Noah: versionning and extensibility came up while reviewing specification that provided or not hooks for extensibility

<Bernard> Review comments can be submitted by sending email to iab@iab.org, or by using TRAC (http://trac.tools.ietf.org/group/iab/trac/report/1 ).

key points are: decentralized extensibility, identification using URIs leading to namespaces, CURIes, etc...

we got also TAG finding about good architecture for document formats, using XML as a tool to demonstrate that, but it was not finished

there are still new things proposed in that space

ht: for example, TAG discussed extensibility in HTML5 and provided feedback

<noah> HT: Extensibility in HTML5 is somewhat circumscribed.

<noah> Also, pertinent to this discussion, HTML5 relies on various forms of registries, including some IANA registries, some proposed Wikis, etc.

larry: there is some similarity in TAG vs W3C and IAB vs IETF, would it be ok to link those efforts between IAB and TAG, like how do you plan to have an influence on IETF

next topic is security.

jar: there are efforts in W3C around security but TAG doesn't have the right expertise to answer security queries. One example is CORS (Cross Origin Resource Sharing), UMP... attemps to turn a browser into a security platform to access resources

<plh> --> http://www.w3.org/TR/cors/ Cross-Origin Resource Sharing

<Bernard> It has also come up in IETF RTCWEB

the TAG looked at that, tried to review it, but it has been a challenge for us. Still we are trying to keep an eye on security issues as it has an impact on ther parts of the Web

<masinter> http://tools.ietf.org/wg/websec/charters

<plh> I would note that W3C has been trying to create a Web Applications Security Working Group for the past 6 months or so

we had a session with Adam Barth on cookies. John Kemp (who left the TAG since then) made good initial work in that area

JonPeterson: we would be happy to hear about documents within W3C that talks about security, would it be good to have a way to refer to documents needing review?

jar: we have a process for that usually Last Call triggers such reviews

ht: it is usually better to use the traditionnal IETF/W3C liaison for that

JonPeterson: the IAB has a broad oversee of the IETF, currently things about identity is of great interest for IAB members

<ht> We will come back to Identity later on

the IAB is trying to look at broader topics that are not addressed by working groups

the IAB once worked on security architecture of the internet, this led to security considerations in groups.

noah: in domains like security, it is difficult to assess things without having deep knowledge of protocols or formats involved

<Bernard> IAB has also held workshops on things like "unwanted traffic", published surveys on DDoS, etc.

<Bernard> http://tools.ietf.org/html/rfc4948

<Bernard> Report from a Security-related Workshop (on unwanted traffic).

<Bernard> Internet Denial of Service Considerations: http://tools.ietf.org/html/rfc4732

ht: the TAG started recently to work at privacy. Both security and privacy issue impact organizaition in multiple ways, and extracting the architecture part of those issues can be tricky

DKA: involve in a serie of workshop on privacy, like device API privacy workshop, as well as the IAB-W3C workshop, the do-not-track workshop.

working on a draft TAG finding relative to data minimization when building an API (Web API)

AlissaC: same situation as the TAG, still in early stages. How do we take guidance to specification writers on privacy topics, trying to do something like what Bernard described for security awareness in groups

<Bernard> http://www.iab.org/activities/workshops/internet-privacy-workshop-2010/minutes/

Ashok: there is a report on the DNT workshop by Thomas Roessler. We are thinking about starting a Working Group on DNT

<masinter> I wondered about whether tracking based on DNS lookups, for example, were in scope, or tracking on SIP telephony

<masinter> i would worry less about drawing boundaries and more about collaborating on the overlap

JonPeterson: we want to avoid a W3C way of doing privacy and an IETF way of doing privacy, so we need to ensure we have the right liaisons

Noah: the TAG doesn't have a directionnal role, we are more alerting on issues

Bernard: how we analyze privacy issues? There are technical issues, but also there are also legal and business issues

privacy is much more complex than security for IETF. ie: what it the threat model here?

<jpeterson> agreed larry

<masinter> will raise venue-shopping during 'next steps' discussion

ht: issue about the conflation between privacy and security. Privacy is in fact far more complicated than protecting your data, so knowing what privacy encompass is important.

boundaries definition is the role of the IETF/W3C liaisons, the TAG doesn't have liaison oversight, unlike the IAB

Larry: would like to hear about what the IAB did wrt liaisons to other organizations

<Bernard> IAB appoints liaisons, shepards liaisons. It is a high level role.

JonP: we have mnot acting as a sheperd for the W3C liaison, he brings regular feedback to us

questions are about what is needed to augment collaborations between the two organizations.

<Bernard> IAB does not duplicate the liaison role.

ht: we forgot to include mnot in this call and it was our mistake.

timbl: there have been issues when html5 evolved as it impacted several groups not only in W3C but also in other organizations.

ht: can you elaborate on 'real time communication' ?

JonP: IETF worked on lots of specification and efforts, now things are moving to the browser, we need to ensure that W3C and IETF side of the efforts, despite the differences between organizational models, need to collaborate more closely.

<Bernard> RTCWEB is a merger of two major technologies: HTML 5 and realtime communications. So it is an inherently complex undertaking.

<Bernard> It will both affect the evolution of realtime, as well as the web security model.

JonP: do you have an opinion as the TAG on where the HTML5 effort is going ?

ht: we are looking at various aspects fo HTML5 and presented our position as feedback to the HTML-WG, some comments were accepted, some other not. But that's the role of the TAG.

<Bernard> RTCWEB will also have privacy impacts. The question is: do we understand how to analyze these issues?

<Bernard> We do not yet have recommendations on how to analyze Privacy Considerations.... security impact is likely to go way beyond protocol-based Security Considerations

<noah> I'm curious how much we know about the likely technical shape of this "real time Web". E.g. what would be the key use cases, protocols and formats

<Bernard> RTCWEB + Geolocation = Emergency Services (see http://tools.ietf.org/html/draft-ietf-ecrit-phonebcp)

Alissa: another impact of RTCWeb is on emergency services, in combination with geolocation. IETF did lots of work in that area, it is not something that could be ignored. Liaison with W3C on that topic is important.

<Zakim> ht, you wanted to ack that the TAG has not looked at RTCWEB as such

<OKolkman> I need to drop. Apologies for that

ht: the TAG has not considered RTCWeb yet, looks like an interesting topic for us

<OKolkman> See you all later Thanks for hosting this call!

<Zakim> noah, you wanted to ask about RTC requirements, specs, etc.

<timbl> http://rtc-web.alvestrand.com/w3c-activity

Noah: wondering what is behind 'RTCWeb', would be interesting to hear form the IAB about requirements and directions of it

ht: IAB and TAG hasn't work together in the past. It is probably an error, and it would be a good thing to have a more 'technical' liaison between the TAG and the IAB

<noah> Thanks Tim. Very interesting. That stuff's all useful, and I'm likely in favor, but it's not the first thing I would have expected from the shorthand "RTC"

<Bernard> See also FCC Location NOI: http://transition.fcc.gov/Daily_Releases/Daily_Business/2010/db0923/FCC-10-177A1.pdf

JonP: we need to look at what our processes are, etc...

<timbl> http://www.w3.org/2001/11/StdLiaison

<Bernard> FCC NG911 NOI: http://www.federalregister.gov/articles/2011/01/13/2011-565/framework-for-next-generation-911-deployment

Larry: we have the issues of 'venue shopping' figuring out which organisation to send work to. Documenting the overlap between organizations would be good. Also our existing liaison could be responsible for identifying topic we could work on together

Plh: we should discuss this, yes

Noah: people doing venue shopping will continue to do so, so such document will not change things, however finding a way to do technical work together is interesting, and documenting that experience after some times would be interesting

<Zakim> ht, you wanted to remember Prague (IETF 80) and look towards the future

<noah> I'm actually going a bit further, saying: let's make sure we take opportunities to actually DO technical work as they arise, and see how it goes.

<Bernard> Example of documenting existing practice in liaison: http://tools.ietf.org/html/rfc4441

ht: the RTCWeb is a good topic to try to do things together

<masinter> update http://tools.ietf.org/html/rfc3205?

<masinter> as architectural issue, how does it apply to RTCWeb

<Bernard> Yes, IAB has discussed potential update to RFC 3205.

<jpeterson> on the iab side of that question ht look at http://tools.ietf.org/html/draft-tschofenig-post-standardization-00

<ht> Jon, yes

<jpeterson> the post-standardization argument is given there

<noah> Curious what IAB knows about its physical meeting schedule next 9-12 months.

<ht> For W3C readers, per that doc't "post-standarization" means standardization, if at all, outside IETF

<masinter> should TAG review that document? Would that be helpful? What impact would IAB like to have on W3C?

<ht> IAB meets at IETF plenaries

<ht> Which are fixed for next 12 mo already

<noah> at...

<noah> e.g. is there one winter/spring?

<jpeterson> ht it doesn't really mean standardizing outside the ietf, it means application protocol development that doesn't require a published specification in any venue because it relies on pushing code

<jpeterson> winter is in taipei

<plh> --> http://www.w3.org/Consortium/Legal/2008/07-W3C-OMA-Agreement.html W3C/OMA MOU liaison

<masinter> http://www.ietf.org/meeting/upcoming.html

<noah> Thanks. Too bad. Very unlikely we could get TAG to Teipei, I'd guess, but we could think about it.

<masinter> November 13-18 Taipei, March 25-30 2012 Paris

<Bernard> Next plenary will cover Privacy at IETF 81.

<timbl> Is there an iCalendar file I can subscribe to for the IETF meetings?

<masinter> tim, http://www.ietf.org/meeting/ietf-meetings.ics

<timbl> tx

Bernard: what is the followup?

Noah: trying to think about that.

Larry: it would be good to have the existing liaisons doing recommendations

Plh: will do

<Bernard> IAB will be at IETF meetings (see http://www.ietf.org/ for current schedule). F2F retreat once a year (date and location not set for 2012).

<Bernard> We can talk offline and see if we can align.

<timbl> Thank you everyone.

<noah> ACTION: Noah to talk to Bernard about possible IAB/TAG co-location [recorded in http://www.w3.org/2011/06/07-tagmem-minutes.html#action01]

<trackbot> Created ACTION-565 - Talk to Bernard about possible IAB/TAG co-location [on Noah Mendelsohn - due 2011-06-14].

<Zakim> DKA_, you wanted to suggest more calls like this might be a good idea as well.

<plh> fyi, the next W3C/IETF liaison call is this Monday

Ashok: we should have more focussed scopes, for example on Security or Privacy

<DKA_> ACTION: Dan to contact Alissa Cooper, organize a future joint discussion on privacy with IAB. [recorded in http://www.w3.org/2011/06/07-tagmem-minutes.html#action02]

<trackbot> Created ACTION-566 - Contact Alissa Cooper, organize a future joint discussion on privacy with IAB. [on Daniel Appelquist - due 2011-06-14].

Noah: would be good to have RTCWeb knowledge, have a telcon on that topic

<masinter> what are architectural issues around RTCWEB that raise it to IAB and should also be brought to TAG

[Switching scribes, now DKA]

Restarting session on Fragment IDs

Noah: 10 minutes. Jeni can you lead the wrapup?

Jeni: larry you were talking about the mime and the web draft?

Larry: The issues raised there seem to fall into 2 categories - interactions between w3c and IANA (extending to other registries)...
... 2nd set: things wrong with the mime registry. To that end, Alexey Melnikov has offered to help co-edit the document.
... so that's what's happening. He suggested we meet together at IETF in Quebec. I will be there.

Noah: Do we care if this becomes an IETF draft or a TAG finding?

Larry: If there's something we can fix through IETF then we wouldn't need a TAG document to endorse these changes.
... looking for some feedback.

Jeni: My feeling is that the level of discussion went into earlier is bigger than the frag identifiers section of that document...
... I think it needs to be expanded quite a lot. Should that be in your document or a separate document?

Larry: I like having a second document better that talks about fragment identifiers.

Jeni: In that case I will take an action to do a draft document that outlines the problems with frag IDs and starts to outline the solutions.

Larry: You could say how things need to change and consider doing an internet draft.

Noah: Are there other actions or clean-up?

[consensus to move on]

<JeniT> ACTION: Jeni to draft a document describing problems around fragids and ways things should be changed due 2011-06-28 [recorded in http://www.w3.org/2011/06/07-tagmem-minutes.html#action03]

<trackbot> Created ACTION-567 - Draft a document describing problems around fragids and ways things should be changed due 2011-06-28 [on Jeni Tennison - due 2011-06-14].

URI Definition Discovery; Metadata Architecture

ISSUE-57?

<trackbot> ISSUE-57 -- Mechanisms for obtaining information about the meaning of a given URI -- open

<trackbot> http://www.w3.org/2001/tag/group/track/issues/57

ISSUE-63?

<trackbot> ISSUE-63 -- Metadata Architecture for the Web -- open

<trackbot> http://www.w3.org/2001/tag/group/track/issues/63

ISSUE-14?

<trackbot> ISSUE-14 -- What is the range of the HTTP dereference function? -- closed

<trackbot> http://www.w3.org/2001/tag/group/track/issues/14

JAR: on Issue-57. History - Tim raised ISSUE-14 in 2002. TAG resolved it in 2005. This percolated through the community. In 2007 we opened ISSUE-57.
... the complaint is that 303s are too slow.
... there are two round-trips required.
... also the complaint that it's hard to deploy - especially for those who use hosted solutions that make it more difficult to send 303 responses.
... I came to this issue in 2007 with a different complaint - I don't know what an "Information Resource" is.
... Harry Halpin has helped - sent an email in 2011 laying out problems with the 303 situation.
... so - how do you use URIs for things other than information resources?
... each URI refers to the information resource that you get at that URI.
... at the last f2f where we discussed it, we agreed I would create a document and then call for public review on www-tag.

http://www.w3.org/2001/tag/awwsw/issue57/20110531/

scribe: we would then decide how to pursue this. We could start a new community group, do a tag finding, charter a new working group, start a document in an existing working group, etc.. [or something else]
... I would like an answer - are we prepared for a consensus process on this issue?
... How I see this issue more generally - this is about the relation of RDF to Web Architecture.
... RDF is supposed to use the URI namespace in a way that is compatible with Web Architecture.
... there is a constituency of people who have problems they need to solve and they feel that Web Architecture is bad for the World. They say the Web Architecture principle as applied to RDF is getting in the way, inhibiting the update of linked data and RDF.
... [adopters could simply choose to go another way.]
... There is also the claim that the specs are weak on how this all works.
... e.g. RFC-2616 bis

Tim: From the perspective of someone implementing a web server, it is strong, but not from a semantic web perspective.

JAR: So - what might happen as this goes to a consensus process?

<Yves> http://tools.ietf.org/wg/httpbis/ 7 drafts

<Yves> http://tools.ietf.org/html/draft-ietf-httpbis-p2-semantics-14#section-8.3.4 for the revised 303 definition I talked about

JAR: Does Webarch have defenders in the linked data community? Tabulator; users of xhv:license, anyone who cares about metadata or provenance...
... most people in the RDF world probably take it for granted.
... having more choice may also be confusing.

Jeni: It's already the case that people trying to do linked data have problems figuring out what to do - to add another story onto that, we make the story worse.
... so [it hinders linked data adoption].

Ashok: If you are able to endorse one story? That would make things better.

Jeni: That also has big implications.

JAR: If there were a REC, what might it say? Strengthen httpRange-14, help out with issue-57; or retract httpRange-14
... that option means that some RDF related working group gets to decide what URIs mean to it -
... this would mean an "opt in" to Web architecture.
... another possibility - if we fail to reach consensus then fall back to a TAG finding, or do nothing. But there is a need for clarity.
... "Undefended linguistic space will be invaded."

<Ashok> Jar: I've told you my take on this ... do you want to give me advice

<Ashok> ... the effect may be that a community secedes from Web Arch

<Ashok>LMM: I didn't understand the stuff in sections 3 and 4

Larry: can we bring more light to bear on the choices?

JAR: Probably that's why I'm asking for review.

[document review]

<masinter> I want to understand the idea in HTTPRange-14 that a resource is either an information resource or not, and whether that can change over time... and the substantive time-dependency makes me not like range-14

http://www.w3.org/2001/tag/awwsw/issue57/20110531/

Ashok: Can you tell me how this is different from the other use case on Metadata?
... The Mark Nottingham use case?

JAR: This is where you're trying to extend the dictionary.

http://www.w3.org/2001/tag/awwsw/issue57/20110531/#id35291

<masinter> the dictionary has a definition of the word. The word doesn't have the definition in the dictionary. There are multiple dictionaries, which have different definitions.

Tim: Word and URI are in the same [category] - a collective noun.

Noah: [thinks URIs are different category than words]

[ moving on ]

<noah> Would "establishing the referent of a URI" be closer to correct than "defining a URI"?

<noah> I think words and URIs are different in that the word dog existed and was used to refer to a class of animal long before there was writing, and thus before the 3 letter string "d-o-g" was used to represent the word.

<noah> Per RFC 3986, the URI is the string.

Tim: [referencing to 3.5] - the idea of the # is that it binds the URI of the document to a local identifier in any language. Doing this for RDF is really just one small case.
... this was in the Web design before html - a fundamental operation in the theory of Web architecture.

<masinter> i'm concerned about temporal variability (the "thunderstorm") comment

<jar> timbl: 3.5 maybe relay point about abstractness of # operator

<timbl> Some people, when they look at the hash URI system, say things like "Oh, that means the meaning depends on the media type, and that mean that there can be no meaning unless a web access has been given -- and we can't insist people do a web access before uing a name!" or upside-down philosophical confusion like that.

<masinter> i think you need to argue against it without editorializing about it. calling something "upside-down" or "confusion" isn't an argument.

<masinter> my concern about hash and 303 is that the make meaning depend on deployment of services infrastructure, which cost power and money to maintain

<masinter> and that to mean something you shouldn't have to have a foundation

Tim: What is Harry's position?

JAR: His position is the anti-HTTPRange-14 position.

Larry: My concern. I want to say something today and mean something 100 years from now without having to endow a foundation to maintain that URI. I would have to guarantee that 100 years from now that there's something there to return 303s.

Tim: You're confusing 2 things. Supposing you want to use URNs. In the URN it's perfectly reasonable to give the URN for the american citizenship act 2012 with a #citizen - to mean citizen as defined in that document.
... this is very common.
... I think the URN issue and the hash issue are orthogonal.

<masinter> i'd like to see in section 5 the issue of requirement for long-term availability of URI data or services (303 or hash)

<timbl> In legal circles, it is very common to define a term with respect to a document such as a law, a la: "resident" in the terms of the Residence Act of 1876.

<JeniT> Is the point that if the semantics of the URI (whether it's IR or NIR) is dependent on what you get as a result of a particular request, you have to still get a response from that request into the future to work out whether the URI represents a IR or NIR?

<masinter> also about the "thunderstorm" example, where the meaning changes if there's a thunderstorm

<timbl> ... That is what the "#" is doing here

<timbl> JeniT, that is a teason I felt HTTPRage-14 was important

<JeniT> s/teason/reason/

<jar> timbl doesn't like postfix #

<jar> ... because "" isn't an ncname

<jar> ... include a warning against using empty fragid

JAR: Rumor: Some tools - site builders - are [munging] the RDFa by cutting off the frag identifiers.

Tim: There's an argument that - "listen - when you're a semantic web company then your world is full of billions of RDF constructs and there is only one document for each one - so the # becomes a pain. We want to hide that. # is an archeological artefact - [vestigial]?"

JAR: Django and Sinatra were the frameworks named.

[some discussion on purl.org reliability issues]

[referncing 3.6.1]

Jeni: Governments would not feel happy using purl.org URIs.

Dan: We also had the same issue with Vodafone.

[ref 3.6.3 - 303s difficult to bookmark - some discussion on lack of ability to copy the previous URI out of the address bar in browsers]

JAR: 303 is "fixed" by HTTPbis - 200 is not fixed by HTTPbis.

http://odontomachus.wordpress.com/2011/03/07/are-you-confused-yet-about-the-word-representation/

<noah> Interesting.RFC 2616 does not use the word "representation" in definition of 200 status code for GET. Rather it says: "the response;"

<noah> So, the question is whether the "corresponding to" relation can hold for other than an information resource.

<noah> I infer that Tim believes: "no, it can't"

Jeni: [on 4.2] if you have a URI that represents schools and a fragment identifier for a particular school, you could return a 404 to say that school does not exist or you could redirect to another document which says it doesn't exist.
... point being you may want to use some of the http status codes to say something about the non-information-resource.

HT: I think 404 always means no representation found.

Noah: [cites chapter and verse of 404 and 410]

[section 4.3]

[discussion on mget]

HT: You might want to reference XRI and a whole bunch of other stuff...

JAR: XRI is using .well-known

Tim: Now things are getting interesting.
... supposing for example - what things could we tweak about the request and the response...

<jar> tweak the content-type

Tim: I think that if people are using an upload space then they can use hashes.

<masinter> i don't think it should cost energy to mean something

<jar> processing instruction ?

Noah: This strikes me as another thing like client side state - we could spend more time on this at the risk of dropping other things?

[break]

rrsagent make logs public

[resume]

[change of venue - Concord]

Tim: "Is a representation of a document" got turned into "representation"
... argument went - HTTPRange-14 was badly phrased.
... then people said the domain of http 200 is information resource - that means you can only use http uris to refer to informaiton resource. Then you would say you can only put information resources on your web server...

Larry: I'm still confused.

JAR: The other document (Web Metadata) is the one that [clarifies].
... people put a document on the web, its title is X so they write RDF that this document's title is X.

Larry: I think it's a problem with the way that it's described and not with the practice.

JAR: People are successfully using this metadata pattern.

Larry: I'm concerned in the way you're descibing the pattern - [dependency on running web server]
... if you do a GET on the string, web server must be running when you do that GET.

Tim: There's a fundamental switch - http is a namespace - you use it for naming things. One property is - you own domain names so you can allocate names to things. These names have a second property that on a good day you can find out stuff about them.

Larry: I understand that but I'm concerned about how it's described.

JAR: I described it precisely. Based on what http-bis says and what people already do...
... 2 documents - the information resource document - a distillation of a [long] process. Stands on its own, independent of issue-57. It says you could go either way on this.

Ashok: I think you got to tell the world what you would like to happen.

JAR: I don't know what I'd like to happen. I think in this case we can solve this one problem [a threat to linked data].
... half a billion documents deployed using the pattern described in this document and billions and billions more. We want to get it right. [JAR 2011-06-21: This may have been a reference to the volume of use of the two patterns, URI meaning its document vs. URI meaning something that some document describes]

Larry: asking about sometimes CC license is embedded...

HT: Sometimes that's not allowed eg for image formats.

JAR: That's right - that's why we need this.

HT: How can we turn this into a decision process? We need to get to a bottom line.
... we need a story of how to get there from here.
... of the possible ways forward, what are the ones we can get to from here in a way that's believable?
... e.g. "solutions that require changes to the server are impossible because we can't get them done."
... if in fact that's only relevant in the corner cases and we could tell the new story from day 1 because the 99% case you don't have to worry about.
... suppose this straw man: suppose we endorse the "pun"?

Tim: There's a third way - Sandro's [work]...

HT: ... but - the problem with the pun is that there are URIs out there that you want to use in both ways - both to refer to the document and what it refers to....

Tim: In fact, we need to ground this. What exactly should implementers do?
... could we create a new way to do this? Create 209?

JAR: The one I tend to favor right now is the Pun solution.

[this is 4.4]

Tim: There's a vast number of cases of web pages about web pages - that this can't handle.

JAR: I've come to believe that these cases are fewer than you think they are.
... Well - it's all of FlickR (for instance).
... now -we could go to FlickR and ask them to change. But I don't want to do that until we have a spec.

Larry: if you believe that linking to an image by embedding is a copyright infringement then...

JAR: Yes but [poem in the comment] case. - [in this case you would draw the conclusion that the CC license applies to the poem which is not the case].

Larry: in the case where the book is by john but bill wrote the preference...
... in defence of punning. You're asserting that this metadata that you're applying belongs to the thing and the document about the thing...

JAR: In the case of license, this is a long conversation.

Tim: When you said - if it works for most of what's out there now that's really important. Yes - it should work for semantic web stuff, but it should [also] work for The Web.
... URIs identify Web pages.

HT: Linked data people are not interested in data about documents.

Jeni: Some aren't.

JAR: We can ask Harry what he thinks.

HT: I was going to take the use case - consider the library of congress catalog entry for Moby Dick.
... and suppose consensus emerges that we have one URI to talk about Moby Dick. You then say that the author is Herman Melville using dc:author, etc...

and other namespaces...

scribe: but they're not good because now I need to say that the author of the card catalog entry is so-and-so....
... then you're not going to be able to use creator.

Tim: Parallel properties - in the open graph protocol on Facebook you get a page about a movie. The RDF references "" (empty URI) is of type "Movie" etc... Sandro says - since they use their own namespace ogp: then we know that ogp: usage means that the referent is the principal subject of the resource referenced.

JAR: That is the Pun solution.

Noah: The complication here is that there are some properties you only want to have one name of but you define parallel properties - one for the author of the page, and one to the principal subject.

DKA: But then you need to have a list of all properties somewhere?

Jeni: No.

Tim: This X is a page about this. dc:author is the author of the page. ogp:author is the author of the principal subject.
... you can say there are relationships between these properties.

JAR: You can say this is the license of this, and you can also say this is the license of what this is about...

Tim: The number of levels of indirection is in the eyes of the beholder.

JAR: THis is a pattern for complying with HTTPrange-14.
... but FlickR would need to be changed.

Jeni: you would need a new property in the case of FlickR.

HT: The real win would be if I could still use dc:creator if there was only one possible interpretation given the subject. If I say the dc:creator for this web page (a bit of ascii art) is Henry Thompson. Then there's no question that creator is being used about the web page. On the other hand if I use it about a URI that doesn't resolve...
... it can't depend on doesn't resolve. What I just said was wrong.

JAR: what's the decision procedure?

HT: We have to have another go at this ourselves

LM: We have a provenance group

JAR: The provenance group is asking us for help

LM: Do the provenance requirements help us to make a decision?

JAR: They invited me to their meeting

Yves: July 6-7th at the Stata centre

[discussion on meta-meta-data]

Summary of Action Items

[NEW] ACTION: Dan to contact Alissa Cooper, organize a future joint discussion on privacy with IAB. [recorded in http://www.w3.org/2011/06/07-tagmem-minutes.html#action02]
[NEW] ACTION: Jeni to draft a document describing problems around fragids and ways things should be changed due 2011-06-28 [recorded in http://www.w3.org/2011/06/07-tagmem-minutes.html#action03]
[NEW] ACTION: Noah to talk to Bernard about possible IAB/TAG co-location [recorded in http://www.w3.org/2011/06/07-tagmem-minutes.html#action01]
 
[End of minutes]

Minutes formatted by David Booth's scribe.perl version 1.136 (CVS log)
$Date: 2011/06/21 14:24:51 $