Re: Notes from Oct 23 IETF/W3C coordination call

[Leslie Daigle <[e-mail]>]

A comment from someone who was not there -- it would be
valuable to list the attendees of the meeting (particularly
as the notes reference various people by first name only).

WRT process -- W3C will look after posting in their
space, and you can submit them as a "liaison report"
next month when the IAB asks for same.  They will,
eventually, theoretically, appear somewhere on the IAB
website in public and findable fashion.

Leslie.

Lisa Dusseault wrote:
> 
> These notes are reconstructed from the Jabber transcripts.  Thanks 
> Thomas and Mark for taking the jabber notes.  
> 
> Please send comments by Nov 13 so I can get these out (and by the way, 
> what's the process for that?)
> 
> Lisa
> 
> ----- 
> 
> Agenda (after bashing on call): 
> 
> 1.  Work done by Content Transformation Task Force in the Mobile Web 
> Initiative Best Practices Working Group -- Dominique Hazael-Massieux to 
> join the call -- 15 min
> 2.  HTTP BIS WG -- Lisa - 5 min
> 3.  Update on the recently-held W3C Workshop on XML Signature and 
> Encryption -- Thomas -- 5 min
> 4.  Update of WS-Policy and SparQL media type -- 5 min
> 5. Broader BCP 56 discussion -- everybody - 15 min
> 6.  Schemes and Protocols -- 
> http://www.w3.org/2001/tag/doc/SchemeProtocols.html -- everybody --  15 min
> 7.  Agenda for Vancouver IETF -- should these topics be on agendas for 
> more discussion? -- Lisa
> 8.  Content t\ype in HTML 5, link rels
> 
> 1  Content Transformation
> 
> New work in http://www.w3.org/2005/MWI/BPWG/  looking at existing 
> technologies
>  - based on existing W3C TAG findings
>  - working on issues like, how to make sure that a page that is already 
> transformed isn't transformed again, how to express transformation policy;
>  - note recent outcry in UK, because vodafone transcoding proxy messed 
> up existing mobile web sites
> 
> Discussion of overlap with OPES: since OPES also has transforming 
> proxies, in this case HTTP intermediaries.  Does new WG have same 
> requirements for disclosure?   IETF/Apps wouldn't force anybody to do 
> things this exact way, but there may be lessons to be learned.  For 
> example, why has there been no uptake?  Apps ADs aren't aware of flaws 
> in the technology, so it may be a socio-economic reason why there's no 
> uptake (e.g. the people in a position to implement OPES aren't the 
> people who want it)
> 
> Announcements requested to go to: apps-review, HTTP WG
> 
> 2  HTTPbis WG
> draft charter: 
> http://lists.w3.org/Archives/Team/w3c-ietf-coord/2007Oct/0001.html
> Charter: http://www.ietf.org/html.charters/httpbis-charter.html
> 
> Mention of call from semweb community to see what HTTP codes mean.  The 
> definitions in the standard are sometimes ambiguous and some things 
> aren't covered. Can the HTTP WG crisp up those definitions?  Lisa 
> responded that this kind of effort, crisping up definitions of things 
> that have already been standardized, seems to result in years worth of 
> bitter arguments (e.g. discussion of what a lock was in WebDAV; 
> arguments on resource instances occasioned by Mogul paper 
> http://www2002.org/CDROM/refereed/444/).  HTTP has many such problems, 
> see similar issues with resource, response, entity.  Conclusion was that 
> for the semweb community to bring that up in the wider community at this 
> point would mean more argument and no benefit.  Somebody suggested to 
> produce OWL ontology to do formal model and maybe tie to bits over the wire.
> 
> Question about charter -- are test suites being developed?  The IETF, 
> like W3C, doesn't do compliance outside specs.  The W3C, however, keeps 
> test suites.  We think this is important for HTTP to crisp things up.  
> One thing the IETF has done is to produce an I-D or RFC that shows 
> torture tests and the like: e.g. request and response messages in text 
> form.  Producing the software of a test suite has never been a IETF WG 
> deliverable, but it has been done closely with WGs.  Similarly, Interop 
> events never done officially by an IETF WG, but in concert with one
> 
> Patent policy for HTTP?  Tim points out that there's an effective 
> royalty-free zone for some standards, it's awkward that HTTP isn't in 
> there.   Lisa adds that there's no guarantee that HTTP isn't in the 
> royalty-free zone; we just don't know because it was developed with a 
> lot of things left unsaid about patent status.  In theory an IETF WG can 
> include technology that is patented along with licensing assurances.  
> However, since the WG is revising HTTP without adding new inventions, 
> it's unlikely that the WG can add anything that's known to be patented.  
> So HTTP's patent situation will likely remain the same: muddy, but no 
> known problems.  We don't see a concrete action to take at this point.
> 
> NOTE that participants in HTTP WG have to disclose relevant patents even 
> if the patents are not new. 
> 
> 3.  XMLsec workshop
> 
> Workshop report releasing today: http://www.w3.org/2007/xmlsec/ws/report
> Main outcome -- there is some work to be done, both on the crypto side 
> and xml side.  IETF involvement is coordination.  Thomas will send 
> pointers to SAAG.
> 
> Thomas would like IETF feedback on the path forward for selecting hash 
> algorithms (e.g., sha-1). Also, what the IETF is (not) doing with crypto 
> suite B.  The  TLS WG has been working on revisions to the TLS spec for 
> adding cyphersuites, and particularly for hash agility.
> 
> 4.  WS-Policy and SPARQL media types
> 
> Thomas sent e-mail about registering thes, was told to contact IANA; 
> they came back and said to talk to the IESG.  Chris agreed to get that 
> on the IESG agenda.  
> 
> 5/6.  BCP56, schemes and protocols
> 
> Issues about BCP56 have been coming up.  BCP56 was written with the 
> experience of abuses of SMTP and other e-mail protocols overloading a 
> single port.  It would be bad if use of port 80 required firewalls and 
> other intermediaries to get deep into the protocol to be able to ensure 
> security or other services. On the other hand, we know that BCP56 isn't 
> necessarily the best advice we could give today.   
> 
> A new port example is IPP:  you can get at IPP resources with HTTP, 
> but conventionally they're named ipp:// and not on port 80.  However, 
> there are more examples, and more recently, which do not use a new 
> scheme nor a new port.  People want new protocols to be easy to 
> implemen.  Atompub WG explained issue quite succintly -- want to be able 
> to write atompub servers with java servlets or ruby on rails, without 
> touching HTTP server engine.
> 
> Beyond this formal standards work, there are a huge number of ad-hoc 
> protocols.  We can't put all these on separate ports or schemes (or can 
> we?)  
> 
> Another possibility, and perhaps more realistic, is to use separate 
> content-types.  The content-type in the client request serves two 
> functions, of explaining how to interpret the client request (schema 
> and/or language) and also to request a particular service of the 
> server.  For example, some certificate retrieval stuff works this way, 
> where a MIME type for the request advertises that the request contains a 
> detailed query and the response should include one or more certificates.
> 
> Is separate Content-Types is also a can of worms?  Does it amount 
> to content negotiation for format which is mostly considered a failure?  
> One doesn't need to see this as negotiation, but rather as an indication 
> what the application is.  
> 
> If most people just use application/xml as the Content-Type, then once 
> again we have undifferentiated traffic. Using http scheme, port 80 and 
> POST of application/xml gives nobody reasonable chance at filtering.
> 
> Is the content-type registration system easy enough, extensible enough?  
> Is that why they use application/xml?  Chris points out that registering 
> in vnd.* is easy, and we also have provisional registration.  So why 
> don't inventors of ad-hoc protocols use new MIME types?   Is it because 
> they can't be bothered?  Is the issue not one of simplifying the 
> process, but one of ownership?  Perhaps when a company asks for a 
> registration, they not only publicize their intent, they also tie 
> themselves down to one thing (one MIME type, scheme, or whatever) and 
> this is the basic reason people make excuses -- they just don't want to 
> limit their own actions like this.
> 
> Topic: Next meeting
> 
> Prefer February: Philippe Le Hegaret will get a date selected. 
> 
> 

-- 

-------------------------------------------------------------------
"Reality:
      Yours to discover."
                                 -- ThinkingCat
Leslie Daigle
[e-mail]
-------------------------------------------------------------------