XML Processing Model WG

Meeting 191, 14 Apr 2011


See also: IRC log


Norm, Paul, Vojtech, Jim, Henry, Mohamed [IRC only], Alex


Accept this agenda?

-> http://www.w3.org/XML/XProc/2011/04/14-agenda.html


Accept minutes from the previous meeting?

-> http://www.w3.org/XML/XProc/2011/02/24-minutes.html


Next meeting: telcon, 21 Apr 2011?

No regrets heard

Meeting at TPAC2011

Norm: I have asked for a meeting slot; I've pencilled us in for a f2f meeting during TPAC 2011, the first week of November in Santa Clara, CA, US.
... Anyone know for sure their plans?

Paul: I expect to be there.

<ht> I hope to be, yes

Norm: So do I.

Jim: Probably not.

Vojtech: If there's work to do, then I can arrange something.

Norm: Ok, the onus is on me to get an agenda together int ime for you to make that decision.

Maps in XPath 3.0?

Norm: Do we have an opinion about maps?

Henry attempts to summarize the state of maps.

<jfuller> I like maps

Norm attempts as well.

Norm: In XProc 1.0, we only have strings and only XML data flows between steps so it's not clear how we would use them.

Vojtech: I had to implement maps as extension steps for some internal use cases.
... I keep them as a global map because they were done as steps.
... I think it would be good to have them, but it can be solved in an implementation-dependent way.

Norm: I think if there is an XProc 2.0, then one of the things we might do is relax the restriction on variables. At that point, we probably want maps.

Henry: I've done some explorations in this space and looked into the most efficient implementation. I think that's a good reason to have maps as a first class data structure.
... If I, as a user, have a map with 10000 elements, I shouldn't have to worry about the most efficient way to update it. That's the implementation's job.

Some discussion of mutability. Maps aren't mutable.

<ht> I endorse John Snelson's observation wrt copying

Norm: It sounds like we'd support the idea of maps, but we can't use them in 1.0.

<ht> +1

Norm: Anyone object to me telling the XSLT and XQuery WGs that?

No objections heard.

Jim: XML serialization?

Norm: Yes, that's possible.

Jim: What about QNames?

Norm: There's a literal syntax for those too.

Last call of XML processor profiles

Norm: Thank you Henry for doing all the heavy lifting in getting that spec out.

Henry: No problem, sorry it took so long.

Norm: No worries.

Recommended processor profile in the browser

Alex: I've been looking at the browser XML processor, specifically in WebKit.
... The browsers don't like to go fetch external resources unless they absolutely have to.
... In the case of HTML, you have to go get scripts and things.
... But in the case of XML, that's not necessarily the case. In WebKit, they've just turned off processing external entities.
... If we don't do that, then we can't do the Recommended profile. So which profile can I follow?
... Then I realized that I lose XInclude and that's something that I think would be useful to have in the browser.
... So I feel like I don't have a profile that fits a web browser where you don't go get external definitions but you would like XInclude.
... Lots of languages these days aren't defined in terms of DTDs, so maybe it's a mistake to have them.
... I looked a little bit at MathML, because they need entities and that's the main reason for the external subset. MathML 3 seems to just say "just use Unicode".
... So I have a feeling we have a mismatch.
... Henry, did you have specific things in mind besides math?

Henry: I don't want to lose it as a profile, because it continues to be the profile I want. I wish the browsers would implement it.
... When I develop with standard DTDs, I can't just hand the result to the browser, it's a real pain.

Alex: I hear what you're saying.
... If you use DTDs, you don't get the behavior you want. The real problem is how HTML is processed which doesn't need DTD processing.
... XHTML is an XML language, and they don't want to go get external declarations in that case.
... It turns out to be really complicated to get external declarations for XML but not for XHTML. So people just say they don't want to deal with DTDs.
... I'm not sure what to say there.

Henry: Neither am I.
... I not also John Cowan's comments that recently came in. He doesn't like the last one either.
... He doesn't like the names full stop.

Norm: If all we had to do was change the names, that'd be lovely.
... I don't know what to do about the browser case.

Henry: Going back to way back to one (but not the only) item that's near to the director's heart is the question of what infoset the author is committed to.
... And my feeling is that it's pretty clear that its the last one and only the last one. Crucially, if i have a DTD in the document and in the external subset I define parity as a general entity who's value is "not" and I write "I do &parity;(insert inflamatory reference)" in a document, to what is the author committed?
... Clearly it's not the version of the document that doesn't have the value for the entity.

Norm: Yeah. I think that's a pretty compelling argument for not "recommending" anything that doesn't do the external subset.

Alex: Maybe we need to add something to deal with the standalone declaration.

Henry: Bad idea. No one understands it, no one uses it correctly.

Norm: I expect Michael Sperberg-McQueen to file a comment about that, based on conversations we had in Prague.

Alex: It does answer the question from the author's perspective.

Norm: It's interesting, could we "recommend" only using documents that have standalone=yes.

Henry: On the web, we could. I don't know if I want to.
... The other side of that is the XML promise, that all XML processors can process all well-formed documents. Surely if we recommend standalone=yes, browsers should reject documents that assert standalone=no

Norm: So where are we?

Alex: It's a real issue because external subsets become a bottleneck.
... We should have good, solid answers to questions about how you deal with the questions.
... And if I fall back from the recommended profile then I lose XInclude which I don't want to lose.
... I wonder why we don't have one inbetween?

Henry: Basically because we thought five was too many.

Norm: I'm inclined to give this a week.

Henry: Can I address a completely different issue wrt the spec?
... To call people's attention to the fact that I restructured things a bit to try to make the relationships clearer.
... This was in direct response to a comment from Liam just as we were going out the door.
... If anyone has any problems or spots any errors, it would be helpful to hear about that.

Norm: I like the fact they're links now.

Any other business?

None heard.


Summary of Action Items

[End of minutes]

Minutes formatted by David Booth's scribe.perl version 1.135 (CVS log)
$Date: 2011/04/16 21:28:57 $