See also: IRC log
2006-08-14 minutes approved without objection
<scribe> chair: highlights lack of participation in the testsuite
<scribe> chair: cr33 discussion is time limited to 45 minutes
Dug: outlines proposal http://lists.w3.org/Archives/Public/public-ws-addressing/2006Aug/0078.html
Anish: other editorial nits, e.g. changing section naming
<scribe> chair: what's the impact on our namespace?
Dug: don't think it impacts it
<scribe> Chair: moving from CR to PR, are these implementation impactive?
Anish: we have a policy,
<<are we backwards compatible>>?
... we're replacing wsa:Anonymous with wsa:NewConnection
<marc> I think it is a breaking change and we should change the wsdl namespace
Anish: seems like a breaking change
David_Illsley: does that bounce us back to CR?
Anish: depends upon the implementers / implementations
bob: I think this changes the WSDL namespace
<TRutt_> It does change the wsdl namespace in my opinion
<chair> chair: will talk to the team to understand the impact from a process POV
<scribe> chair: what's do folks think about the proposal
marc: my concern is to keep ReliableMessaging dependency / text out of the spec as WS-RX isn't baked
Anish: RM is in public review, any reference to that document will be stable, and the concrete example helps clarify and adds perspective
dhull: likes the approach of the proposal over the status quo
David_Illsley: supports the RM reference
Tony: we still don't have a good definition of the anonymousness of the anonymous and reliable addressing URIs
Dug: RM spec defines that
Tony: but my implementation may be RM ignorant
Dug: proposal doesn't address that issue, rather enables RM to layer addressing
Tony: i think we should address that issue
Anish: the proposal makes wsaw:NewConnection element talk about cases where you have to establish new connections or not, and opens it up for extensibility, if you don't understand the URI (the spec that defines it) then your not going to understand and know what to do with it
<TRutt_> where is reference to wsrm in the proposal?
<dhull> +1 to not referencing RM (are we?)
<dhull> (i.e., are we in any substantive way)
mrGoodner: Anish's point makes me question the reference to RM, and makes me uncomfortable, I'm not sure it's a service to the reader. What sounded like an editorial issue
<David_Illsley> the term addressable EPR refers to the ability to initiate a new connection to that EPR. Examples of non-addressable EPRs are EPRs containing â€œhttp://www.w3.org/2005/08/addressing/anonymousâ€ as the value of wsa:Address, EPRs containing a URI that matches the URI template defined in WS-ReliableMessaging Section 3.7 [informative ref] as the value of wsa:Address
pauld: hates the idea of referencing RM, that would be very wrong, hate the idea of allowing other anonymous URIs and having to understand other specs to realise the semantics wrt to backchannels
Anish: any reference to RM would be just as an example
dhull: what's a connection?
Dug: initiating a new connection is used within the core spec
dhull: used, perhaps, but unlikely to be defined
<anish> connection is talked about in the soap binding not the core
<dhull> It's specific to SOAP 1.1/HTTP: When "http://www.w3.org/2005/08/addressing/anonymous" is not specified for the response endpoint, then the message SHOULD be part of a binding that supports not returning a SOAP envelope in the HTTP response (e.g. see [SOAP 1.1 Request Optional Response HTTP Binding]). Any response message SHOULD be sent using a separate connection and using the address value...
<dhull> ...specified by response endpoint. Note that other specifications MAY define special URIs that have other behaviors (similar to the anonymous URI).
marc: you need to pick your URIs carefully
pauld: that's my issue, OASIS site can expect hammer from implementations which understand addressing but don't know or care about RM
Dug: that issue exists anyway, people will check WSDL first
pauld: that assumes you have a WSDL
dhull: dangerous to rely upon people being smart
<bob> +1 to dhull's warning about depending upon people being smart
dhull: i worried about using the opening new connection terminology, ok for HTTP, but there are other protocols for which it may not make sense, esp within the terms being discussed in XMLP for one-way SOAP MEPs
Anish: question for Marc: let's say this element exists and is defined in WSDL but for some reason the receiver doesn't understand RM, and compare that with the case without this proposal and you get the RM anonymous URI, then you're still going to [open that socket to the OASIS site] it doesn't make things worse
dhull: difference between addressable and non-addressable EPR, and it depends upon the binding, the exceptions are indepenedent of the binding, and the anonymous sits between these two camps, we can improve the quality of the spec by working on sharpening these concepts, may be an XMLP issue
<dhull> yep ... the SOAP specs are are only well-defined interface to bindings
paco: use of anonymous in the SOAP binding document, anonymous was pegged to a particular message in the SOAP MEP, maybe using XMLP terminology in Dug and Anish's proposal may help
dhull: maybe taking about separate MEPs as opposed to connections?
bob: would such a change to the proposal, as well as loosening the reference to RM remove any objections?
Tom: this proposal with the clarifications, does that mean we have to change the anonymous URI, can't we just change the semantics?
Anish: yes, we've found anonymous confusing, so would like to remove the term
Dug: if that's the breakage, and not removing the word "anonymous" helps, but would prefer to remove it
pauld: not going to lie down in the road over decisions made last week, but stills feel strongly over not referencing WS-RM
<anish> how about 'SameConnection'
Tony: agrees with Tom in keeping backwards compatible, but still worries about how to tell if a URI is anonymous
Anish: how about "same connection"?
omnes: sounds interesting
<David_Illsley> SameConnection doesn't sound right when you're using SOAP/JMS with the request-response MEP
Tom: why do you have to annotate the WSDL when you are enabling RX
Dug: maybe a conflict between the wsaw: anonymous and any spec layering on top of such a WSDL
MrGoodner: sympathise with Tom's point, RM does allow that make connection to be made, unclear on the concern from RM in this case
Dug: can't put optional on anonymous as that means you could use a concrete address as well as another anonymous URI
dhull: explains his position made in IRC, much of this discussion is out of scope for this WG
paco: this is a core wsa concern,
wsa enables async behaviour, so isn't out of scope
... section 5.1 does enable the WS-RM behaviour, we should look at this in some detail
dhull: meant "out of scope" we
don't have the tools to deal with this
... we have colloquial terms, but nothing more precise
bob: time's up!
... well almost
... how about MUST to a SHOULD?
... would that release fewer worms?
Tony: SHOULD irks me, opens the door to other possibilities
bob: what's the shortest path? folks could come back with better wording, another proposal, or we could close with no action?
Tom: if the endpoint acknowledged RM was in use, it could extend what anonymous should mean. That's what's not properly addressed in the proposal. It's a composition thing
bob: 5:03 eastern, let's wave a fond fairwell to Dug
any objections to the errata?
RESOLUTION: close CR-27 with Philippe's propsal
Tony: has a single sentence proposal
<MrGoodner> gotta go... bye
Anish: is 'empty' a special case
Tony: you can't make wsa:Action empty
Anish: it would be the default
... we need an exception for empty
RESOLUTION: close CR-30 with Tony's proposed text + exception for empty IRI as a valid SOAPAction value
<dhull> the old gray matter ain't what she used to be
Tony: outlines his proposal, Tony rules verus Jonathan rules .. http://lists.w3.org/Archives/Public/public-ws-addressing/2006Aug/0083.html
David_Illsley: language isn't
stronger than "MAY", anything can happen
... especially in CHANGE 3
discussion on CHANGE 3, too loose
Tony: prefers CHANGE 2
bob: objections to accepting CHANGES 1 &2
<scribe> ACTION: Tony to implement CHANGE 1&2 to the table in preparation for CR-31 [recorded in http://www.w3.org/2006/08/21-ws-addr-minutes.html#action01]
bob: weekend after next is Labour day
pauld: August 28th is a public
holiday in the UK, i won't be there
... but is available on the 4th of September