See also: IRC log
-> http://www.w3.org/XML/XProc/2012/05/31-agenda
Accepted.
-> http://www.w3.org/XML/XProc/2012/05/24-minutes
Accepted.
Regrets from Vojtech
<scribe> No progress
-> http://www.w3.org/XML/XProc/docs/langreq-v2.html
Norm: Murray, what can we do to be helpful?
Murray: Good question.
... It's a big document and there's a lot here, but the
organization may need work.
... Let's start with variables and debugging on p:log.
... What information can we get out of the machine for
debugging purposes.
Norm: I have an action to produce a pipeline that does that.
Murray: Ok, then let's talk about
the resource manager
... I understand that Norm and Cornelia have a resource
manager.
Norm: I do.
Cornelia: Vojtech would know.
Norm outlines his resource manager
Norm: Basically, it caches generated URIs and returns them in favor of any other source if they exist.
Henry: The Markup Pipeline had a
two-part resource manager.
... One part is that the RM made available to steps the ability
to cache infosets or other objects.
... We had compiled schemas, compiled stylesheets, and compiled
pipelines all of which were the results of non-trivial amounts
of processing and were available from the resource
manager.
... The schema validator, for example, before it built a schema
from sources, would check if the RM had the compiled
schema.
... after a staleness check; Norm, you didn't mention that, we
should come back to it.
... But that doesn't seem like something we need to
standardize, it's just an implementation issue.
... The other part was more like what Norm outlined.
... It played a bigger role than what we've talked about today.
It really did serve as a kind of local filesystem for
pipelines.
... In particular, this turned out to be useful in steps that
produce unordered sequences of documents.
... A resource manager might figure into the design for how
steps that produce or consume unordered sequences coordinate on
how the decide what's available.
... At the extreme, a RM gives us a very different processing
model where steps are thought of as producing and consuming
named resources.
... And other steps can use those resources without plumbing
directly between steps.
... A simple way to put that is: what is the impact of having a
resource manager on step dependency calculations.
<scribe> ACTION: Henry to review Markup Pipeline documentation and report anything useful he finds [recorded in http://www.w3.org/2012/05/31-xproc-minutes.html#action01]
Alex: Is the XSLT import/include case relevant here?
Norm: Yes, I think that's exactly the same use case.
Some discussion of caching and resource managers and related issues
Henry: Have we reached agreement
about whether we want to adopt the XSLT/XQuery invariant
... roughly speaking this says that any given XSLT or XQuery
episode does gets on URIs exactly once and you always get the
same document.
Norm: Yes, that's correct.
... But in XProc we didn't want to make that requirement
Some discussion of security
Henry: This is related to the RM
and the ability to attach "out of band" information to
documents.
... We've talked about some possible others, like serialization
parameters
Murray: Are we planning a f2f meeting?
Norm: Yes, at TPAC in Lyon, FR in November, I believe.
Adjourned