XML Processing Model WG

10 Jan 2008


See also: IRC log


Norm, Henry, Paul, Alessandro, Alex, Richard, Andrew



Date: 10 January 2008

<scribe> Meeting: 97

<scribe> Scribe: Norm

<scribe> ScribeNick: Norm

Accept this agenda?

-> http://www.w3.org/XML/XProc/2008/01/10-agenda


Accept minutes from the previous meeting?

-> http://www.w3.org/XML/XProc/2008/01/03-minutes


Next meeting: telcon 17 January 2008?

Face-to-face meeting in 2008?

Norm: Anyone think we want to try to do that before the Tech Plenary? (Oct in Mandelieu)

Henry: I can imagine that we are sufficiently finished with XProc in a few months that it will be time to turn our attention to our other task.
... Sitting around a table would be the best way to start.

Norm wonders about Balisage...

Norm: Let's wait and see

Last call comments

-> http://www.w3.org/XML/XProc/2007/09/lastcall/comments.html

Comment 21: XProc localization

Norm outlines his message about making p:error take its description from an input port

Henry: This resonates with something I encountered when trying to simplify the early example pipelines.
... Sometimes you'd like an alternative between an input port and an option.
... For example, I'd like to be able to specify the XSLT stylesheet by way of an option instead of on an input port.
... It's just laziness, but the idea of having an option and a port in a trading relationship seems very natural
... The alternative is that folks won't provide descriptions for errors.
... Do we want to consider adding this feature?

Richard: Can't you do it with a select and an expression?

Norm: No, because it gets turned into a string and the markup is thrown away.
... I guess my reaction is, gosh that's a nice feature but aren't we done adding features?

Henry: It's not a feature it's a design pattern.

Richard: It seems to me that it's a bit excessive for what it does.

Support does not seem to be rising for the coocurrence constraint idea.

Norm: Do we want to make the description of p:error an input port?
... I observe that it does require the user to make up some random document element for the message.

Richard: What about structured content?

Norm: No, way too much feature for now. We could remove the restrction to strings, but that didn't get support alst time it came up.
... Straw poll: Input or option?

Input wins with unanimity.

<scribe> ACTION: Norm to change the spec to make error descriptions come from an input port on p:error [recorded in http://www.w3.org/2008/01/10-xproc-minutes.html#action01]

Split the spec into two or three parts?

Some discussion of why we might do this

Richard: I'm inclined to keep them together.
... I think it's simpler for readers to ahve it all in one document.

Paul: The biggest advantage as far as I'm concerned is that you could advance different parts to PER, for example, independently.
... For example, you can tell folks we haven't fiddled the language, we've just changed the steps.

Henry made apoint earlier about the fact that breaking it into pieces doesn't necessarily make it into separate RECs

Paul: I don't think it makes sense to do unless we make different RECs.

Norm: The editor doesn't feel strongly one way or the other, for what it's worth.

Henry: I think Paul's argument has value, but I don't feel that strongly either.

Some discussion of adding steps in the future

Henry: At worst, we can always change our minds later.
... At V1.1 time, for example.

The chair doesn't hear consensus for splitting the document at this time.

Replace ignored namespaces with p:appinfo (or some such)

Norm summarizes how we got where we are

Richard: One advantage to inventing your own namespace is that it designates the owner of the info.

Norm: We could mandate, suggest, or leave that problem to implementors

Alex: Why?

Henry: Because it's just not clear and is contradictory in the spec.
... The core of my problem is that we invite people to put an element between two steps but we say it musn't change the flow. That's just bizarre.

Some discussion of the extent to which deleting them is appropriate.

Norm: I guess the position I've come to is that ignored namespaces introduce some complexity without much benefit over having a single element for this purpose.

Alex: What about making them top-level?

Henry: Yes, excpet that the spec tries to do that and it gets confusing.
... Instead of putting it in subpipline, couldn't we put it in the prolog?

Richard: Does having a p:appinfo solve the placement problem?

Some discussion, basically yes.

Alex: I'm partial to the ignored namespace thing, but I don't mind having an appinfo, as long as we don't call it "appinfo".

Straw poll: keep ignored namespaces,or abandon them in favor of some element to be named later.

Results: 6 for new element, 1 for ignored namespaces, and 1 abstention.

Name the new element...



<MoZ> pipeinfo


Results: pipedata: 3, userdata: 2, pipeinfo: 4, pragma: 3

The winner is p:pipeinfo.

Allow sequences on source and result of p:pipeline

Norm: I'm inclined to say not.

Alex: I'm inclined to allow them so that you don't have to jump into a new syntax just to provide two input documents.

Richard: Because this is the simple case, I was expecting it to be used on the command line and that's possibly going to make the output syntax different.

<MoZ> +1 with Richard on output

Richard: I was planning to implement sequences as directories, that means I'll always get a directory even when there's only a single file.
... I can work around it, but it seems a bit yucky.

Some discussion of the various possibilities

Alex: If I put an XSLT inside a pipeline, I shouldn't have to fiddle the syntax depending on whether or not a sequence is produced.

Straw poll: sequence or no sequence?

<ht> I note that saxon 8 just concatenates multiple unnamed result-documents on stdout

Results: sequence: 2, no sequence: 4, abstain: 2

Henry: What about sequence on input, no sequence on output?

Alex: I'm not going to lie down in the road over sequences.

Alessandro: I think we shouldn't let constraints on the command line system influence how we design our language.

Richard: I'd agree, except that we're only discussing the abbreviated form. And that does seem to me to have a more direct connection to useability on the command line.
... It's there for simple cases.

Norm: Is there anyone who can't live with no sequences on input or output

None heard.

Any other business?


Henry: I'd like to encourage the wG to identify any issues between here and last call.

Some discussion of circular imports in pending mail.


Summary of Action Items

[NEW] ACTION: Norm to change the spec to make error descriptions come from an input port on p:error [recorded in http://www.w3.org/2008/01/10-xproc-minutes.html#action01]
[End of minutes]

Minutes formatted by David Booth's scribe.perl version 1.128 (CVS log)
$Date: 2008/01/10 17:04:59 $

Scribe.perl diagnostic output

[Delete this section before finalizing the minutes.]
This is scribe.perl Revision: 1.128  of Date: 2007/02/23 21:38:13  
Check for newer version at http://dev.w3.org/cvsweb/~checkout~/2002/scribe/

Guessing input format: RRSAgent_Text_Format (score 1.00)

Succeeded: s/dot his/do this/
Succeeded: s/ifno/info/
Succeeded: s/n,./n./
Found Scribe: Norm
Inferring ScribeNick: Norm
Found ScribeNick: Norm
Default Present: Norm, Ht, PGrosso, avernet, alexmilowski, richard, Andrew, MoZ
Present: Norm Henry Paul Alessandro Alex Richard Andrew
Regrets: Murray
Agenda: http://www.w3.org/XML/XProc/2008/01/10-agenda
Found Date: 10 Jan 2008
Guessing minutes URL: http://www.w3.org/2008/01/10-xproc-minutes.html
People with action items: norm

[End of scribe.perl diagnostic output]