XML Processing Model WG

Meeting 170, 08 Apr 2010


See also: IRC log


Mohamed, Vojtech, Henry, Norm, Alex
Paul, Murray


Accept this agenda?

-> http://www.w3.org/XML/XProc/2010/04/08-agenda

Mohamed: Don't we get to see comments on the PR?

Norm: Not until the review is over.

Henry: Right. But it's not too late to ask your AC friends to vote!

<ht> http://www.w3.org/2002/09/wbs/33280/xproc/results

[A member-only link]

Henry: It would be good to get some more results to help Ian with the publicity

Accept minutes from the previous meeting?

-> http://www.w3.org/XML/XProc/2010/02/25-minutes


Next meeting: telcon, 15 Apr 2010?

Mohamed gives regrets.

Review of the Default XML Processing Model

Henry: So, Norm and I cooked up this draft and it's received some internal review and I showed it to the TAG

<ht> http://www.w3.org/XML/XProc/docs/defproc.html

<ht> http://www.w3.org/XML/XProc/2009/11/12-minutes.html#item04

<ht> http://www.w3.org/2001/tag/2009/11/06-minutes.html#item08

Henry: XProc discussions focussed on two questions: XInclude fixup, which we decided we wanted to keep, and given that we're no longer talking about this as a default, but rather we're presenting it as "this is something you can refer to". It's not a default, but a sort of preferred or baseline processing model.
... There was some suggestion that we ought to change its name before we publish it.
... The TAG also raised that question. DanC went even further and said "In order to avoid anyone thinking this was the one true model", why don't you define another one.
... I thought that was a suggestion at least worth considering.
... TimBL isn't happy, but I'm not sure we can do anything to make him happy.
... You may recall that the other example that I often referred to was decryption/signature checking. When I returned to this this autumn, I concluded that it didn't make any sense.
... Because 9900/10000 times, decryption involves user interaction. It's bad form to include the keys in a message so that decryption could proceed automatically.
... So with some reluctance, I've taken it out and TimBL would like us to address it.
... Aside from changing the name, and perhaps defining a second model, I think we're ready to ask for FPWD

Alex: Does it make any sense to have more than one model in this document?
... There are some obvious variants that are the next step, like validation.

Henry: The other alternative which I have mixed feelings about is to go the other direction: give a name to the bare minimum.
... No reading of the external subset, no XInclude.

Norm: Does anyone know if you can tell Xerces *not* to read the external subset.

Henry: I don't know.

Alex: Does it matter?

<MoZ> http://xerces.apache.org/xerces-j/features.html

Norm: If modern parsers don't let you do it, then I'm not sure it's good to give it a name.

Henry: I'm of two minds: I don't want to encourage folks to do it, but it is spec-compliant.

Norm: If we want to go that direction, I'd be inclined to make XInclude optional. I don't really want to encourage application authors to do less than read the external subset.

<alexmilowski> http://apache.org/xml/features/nonvalidating/load-external-dtd set to false

Henry: I'd be inclined to keep the first two and get rid of the last two.

Norm: I'd be inclined to keep the first three and lose the last one. Surely xml:id is free?

Henry: I guess, but see your point about whether all parsers support xml:id
... I think, in fact, Xerces rejected a patch to support xml:id

Vojtech: Yes, I think that's right.

<MoZ> https://cwiki.apache.org/jira/browse/XERCESJ-1113

Norm: So the two questions are, do we want to provide more than one, and what should we call the document.
... I guess if we supplied more than one, then something like "Parsing Profiles for XML" might work.

Henry: I still think "processing model" is useful in the title What does the XML spec call this?.

<MoZ> http://www.w3.org/TR/xml/#dt-xml-proc

Alex: Why not The Default XML Pipeline?

Henry: I was thinking we should use the terminology that the language itself uses.
... XML Processor Profiles or something like that.

<MoZ> +1

Norm: I guess that works for me, though I worry that "profiles" is sort of overloaded these days.

<MoZ> XML Processor Level à la CSS

Norm: If we do more than one, then maybe "XML Processing Models" works

Henry: Let's not hold up the discussion for any more discussions about naming.

<scribe> ACTION: Henry to fix typo in the bibliography where XML5e is referred to as XML4e [recorded in http://www.w3.org/2010/04/08-xproc-minutes.html#action01]

Henry: We've discussed at some length doing less as an alternative, there's also a doing more alternative. (1) Leaving it as it is, vs. (2) one or more w/o prejudice to which one.

(1): 0, (2): 5

Alex: Maybe one way to spin this is to divide the document into different kinds of user agents: "web browsers", "web service", "validating authoring tool", etc.

Norm: It's an interesting idea, but are we sure it breaks down along these lines?

Alex: We could qualify it with validation, etc.
... The problem with the document is that it's the "default" model. For whom?

Henry: That's why I think DanC's suggestion is a good one. It'll make the document more useful and more used if we identify several points along the continuum.

Alex: We would just be providing context.

Henry: I'm a little nervous about that. It's likely to only get us enemies.

Norm: I think I'd prefer to define what the pipelines are and let application designers decide which ones to use.

Vojtech: But we have to give them fixed names, so that other specs can point to them.

Norm: Absolutely

Henry: I'm thinking "minimal", "basic", and "validating"

Vojtech: I think they all have to be minimal.

Henry: minimal, the one we have now that's recommended, and one more maybe that does validation.

Vojtech: So folks will add to them. We should have a really minimal one.

Norm: The one dividing line I see is, that there's no point defining pipelines that require additional parameters/options.
... So no XSLT or RELAX NG validation.

Henry: I thought about xml-model and Richard raised xml-stylesheet. They are, after all processing instructions and we're talking about processing models.
... I guess the way to address that is with a few sentences that address those PIs.

Proposal: Let's try to get this to FPWD. I propose we change the name (editor's discretion) and have minimal and basic models.

Norm: Where minimal does 1, 2, and maybe 3. Basic is what we have now.

Alex: So we're not going to say anything about the xml-stylesheet PIs?
... Browsers do that, having it codified as a basic option would be good.

Henry: I think you may very well be right, but I'd like to think about it a bit.

Alex: It would be great to have something to point to that we could say browsers *should* do.

Henry: I see that, but let's get it out the door first.
... What should the short name be?

<MoZ> procmodel

<caribou> I thought it would avoid model in the shortname?

Norm: Let's see what title we get and then figure it out.

<scribe> ACTION: Henry+Norm to have the new draft ready for discussion next week. [recorded in http://www.w3.org/2010/04/08-xproc-minutes.html#action02]

<ht> Carine, I agree wrt model

Any other business?

Alex: Let's get AC reps to vote!

Henry: I think we'll get to Rec w/o any difficulty even if we don't get a lot more votes.

Norm: I think it just makes the press release, media fanfair easier if we have more votes.


Summary of Action Items

[NEW] ACTION: Henry to fix typo in the bibliography where XML5e is referred to as XML4e [recorded in http://www.w3.org/2010/04/08-xproc-minutes.html#action01]
[NEW] ACTION: Henry+Norm to have the new draft ready for discussion next week. [recorded in http://www.w3.org/2010/04/08-xproc-minutes.html#action02]
[End of minutes]

Minutes formatted by David Booth's scribe.perl version 1.135 (CVS log)
$Date: 2010/04/08 15:59:03 $