See also: IRC log
Norm notes that he added XProc errata E10
Henry gives regrets for 11 August and the three following weeks
Jim gives regrets for 11 and 18 August
Norm: I think we'll meet on 11 August anyway and survey the scene. We can agree to start again in September if August looks like a wash.
Norm: Thank you Henry for the new draft.
<scribe> ACTION: Norm to update the comments list and propose an action plan to make progress [recorded in http://www.w3.org/2011/07/21-xproc-minutes.html#action01]
Norm: That will require considering the proposals we've discussed for more radical change in the direction of browser compatibility as well as detailed comments from cmsmcq and others.
Henry: I think it would be useful to discuss what we might do wrt to the browser question.
Norm: I think we need the bottom level of conformance to be what browsers do.
Henry: I don't have a problem
with that, and then lobbying to get HTML5 to reference it. I
think the chances they will are small.
... I don't think we should jettison several of the other profiles to make space for that one.
... The bottom line is that there are an enormous number of XML applications out there that have nothing to do with HTML and do stand to benefit from making use of this spec. And they're more likely to do so.
Norm: I think there was also concern about the word "recommended" in a profile name
Henry: I'm happy to change that, even right now.
Norm: I think another axis of concern was implementing XInclude in the browser. XInclude is only in a profile that reads the external subset and browsers won't.
<Jim> as an aside ... http://www.xml.com/pub/a/2007/03/28/xinclude-processing-in-xslt-with-xipr.html now works with SAXON CE
<Jim> spoke to Erik a few weeks ago in Zurich
Henry observes that reading the external subset w/o validation is useful for attribute types.
Norm counters that attribute types are mostly useful for ID attributes and we have xml:id so why bother.
Norm reads through the profiles suggesting that on reflection the seem pretty good, with the possible wrinkle that you can't do XInclude w/o reading the external subset.
Norm: Saying you do the basic profile with XInclude doesn't seem conceptually that different from saying that you do the modest profile and require validation.
Norm: So maybe we like where we are and we just need to deal with the detailed technical comments.
Henry: Yes, but we probably should rename the "recommended" profile.
Norm: I think "full" is the first adjective that follows "minimal", "basic", and "modest" in my mind.
Jim: Full sounds good to me.
Vojtech: They both sound like something that's good enough. If you want something "more fuller" how would you name it?
Norm: Is there more?
Vojtech: One of cmsmcq's comments
is that this division into four profiles was arbitrary. One
might have a different view. It's more about deciding what goes
into each box.
... From that perspective, full might be bad because it might not be full for someone else.
Henry: We could add some prose in
a suitable place that describes why we named these particular
... but you could define your own based on this pattern.
Vojtech: I think right now the feeling cmsmcq had was that it seems a bit arbitrary.
Henry: Well, the answer is, I
think, that thinking historically about how they came
... There is an inventory of low-level XML specs. And although they are in some sense independent, they are partially ordered.
... You can't really imagine requiring XInclude without xml:base.
... There are two sources of indeterminacy in what you get from a processor. One is how it interprets the flexibility that the XML spec itself provides and the other is which of the low-level specs it supports. The tableaux satisfies the former and the profiles we pick sort of walk up the partial order.
... We decided that there was no point picking the one off the bottom (no namespace support, no xml:base support), and after that I think it falls out pretty straightfowardly.
Vojtech: I don't disagree, I just think the commenter wanted more of that explanation in the specification.
Henry: Would what I proposed help, do you think?
Vojtech: I think it might. We already have some of this in the background section.
Jim: We could also just enumerate them, level 1, level 2, etc.
Vojtech: Not just naming the profiles, but the stuff that goes in them.
Henry: Yes, and I think we can consider this when we look at that comment in more detail.
Norm: So I think my takeaway is we're willing to consider or even anxious to rename the top-level profile if it'll help, but otherwise we should proceed with addressing the technical comments that we got.
Norm: This addresses the issue
Vojtech raised about the discussion of "error" as an "output
... Anyone think I got it wrong?
Jim: Are these errata tied back to email messages on the list?
Norm: No, but they probably should be. At least to the minutes where the decision was made. I'll try to be better about that.
<ht> HST is happy with E10
Jim: What about ancillary specs?
Norm: I have an errata document for the p:template spec. But it's a note so we can just republish it if we want.
Jim: And do you think notes are a good way forward?
Henry: Me too.
Jim: Should we be considering anything else for notes?
Norm: I'm happy to, if anyone has suggestions.
Jim; I think we should be active in publishing those kinds of notes.
Murray: I didn't think the antecedent of "this" was clear.
Murray: Some of the specs refer to external declarations, what about the internal ones?
Henry: The "required of conformant non-validation parsers" clause ties that one down. The spec doesn't give you any leeway there.
Some discussion about the orthogonal nature of validation. You can complement any of these profiles with a statement that validation is required.
Murray: Shouldn't we say why?
Henry: Give my regards to all and sundry in Montreal.