See also: IRC log
http://www.w3.org/XML/XProc/2007/10/25-agenda
Accepted as distributed
http://www.w3.org/XML/XProc/2007/10/18-minutes.html
Accepted as distributed
End of summer time in Europe
Agenda wrong [now corrected]
Call next week remains at 1100EDT and related times in the US, but for that week only is 1500GMT for the UK and 1600CET for the rest of Europe.
Regrets from Norm Walsh, Paul Grosso, Rui Lopes for 1 Nov
A-86-04: Done
A-87-01: Done
Rest are continued
http://www.w3.org/XML/XProc/2007/09/lastcall/comments#C029
http://lists.w3.org/Archives/Public/public-xml-processing-model-wg/2007Oct/0084.html
HST: [Introduces the proposal in the above email]
AV: Doesn't this introduce an
unpredictability?
... It's not clear which case you're in
... You might leave a declaration out when you needed one
RT: If that happens, an error will always result, because there will be an unconnected primary output
AV: I'm thinking about someone
reading a pipeline and trying to tell whether there's an
output
... they have a hard job
AM: If someone uses the default form, readers will have to understand the default form
HST: AV, what do you recommend?
AV: I don't mind always having declarations on pipelines
MM: Richard, could you summarize?
RT: It's annoying to have to
write declarations for simple one step after another
pipelines
... You should be able to just wrap a sequence of steps in
p:pipeline and have a runnable pipe
... The analogy is to UN*X pipes
MM: With stdin and stdout
RT: Right
MM: And the problem?
RT: Well, the way we implement
this is by asking if the last step has a primary output, in
which case it gets hooked up as the output of the
pipeline
... but if the last step is itself a pipeline, or a choose,
things get messy
... HST's solution says in the pipeline case, it has to have
declaration, so you don't have to start the process all over
again
HST: Propose straw poll: 1) Adopt HST's proposal; 2) Revert output declaration defaulting on all compound steps; 3) Revert output declaration defaulting on p:pipeline
Prefer 1) HST, 1/2RT, AM, RL; 2) ; 3) 1/2RT, AF, AV
MM, PG preferring to try harder
Can't live with: 1) ; 2) HST, AM, AV, RL, AF, MM; 3) ;
3 1/2 for (1), 2 1/2 for (3), so I'm not going to call the question
Please add discussion by email
http://www.w3.org/XML/XProc/2007/09/lastcall/comments#C006
http://lists.w3.org/Archives/Public/public-xml-processing-model-wg/2007Oct/0075.html
HST: [HST summarizes]
HST: I'm confused -- how can you put p:pipe in something static such as a step type definition
RT: The discussion started wrt p:pipeline, are we extending it to p:define-step?
HST: The prose in the spec at the moment doesn't distinguish this case
RT: I don't think p:pipe makes
any sense in the case of p:declare-step
... Also, there's an interaction with default output
declarations, as regards whether the output port is bound or
not
... if we allow p:pipe in default bindings, the above may
depend on whether or not the relevant step has input, which
is surely too complex to manage
AM: Can't we just get rid of this?
HST: Absolutely right
... we have no use case, let's get rid of it
RT: I believe the status quo does not allow a 'default' binding for p:input in p:declare-step
HST: Agreed
RT: Not clear whether Norm's proposal is only for p:pipeline, or for any declaration. . .
HST: [works through thought experiment, worried about deadly embrace]
RT: Don't see any problem
MM: What's the problem doing this just on p:pipeline
HST: I guess I see no contradiction, I thought there was. . .
PROPOSAL: Adopt the solution outlined in http://lists.w3.org/Archives/Public/public-xml-processing-model-wg/2007Oct/0075.html, for p:pipeline only
AV: Does that leave the question open for p:declare-step?
HST: No, it doesn't, it rules it
out
... Convinced to withdraw the above -- declare-step will have
to be reconsidered
RT: Is this top-level only? Or pipelines in libraries?
PROPOSAL: Adopt the solution outlined in http://lists.w3.org/Archives/Public/public-xml-processing-model-wg/2007Oct/0075.html, for top-level p:pipeline only, leaving p:pipeline in libraries and p:declare-step open
RESOLUTION: Adopt the solution outlined in http://lists.w3.org/Archives/Public/public-xml-processing-model-wg/2007Oct/0075.html, for top-level p:pipeline only, leaving p:pipeline in libraries and p:declare-step open
PG: HST got it wrong -- the
meeting remains in EDT, and is only changing wrt Europe, which
is changing from CEST to CET
... and Britain, which is changing from BST to GMT
No new actions.
[End of minutes]