Meeting minutes
Accept the minutes of the previous meeting
Accepted.
Review of open actions
No progress reported.
Status reports
Bug reports and fixes here and there.
New open issues
None.
Pragma requirements
Bethan: Thanks to everyone who sent me requirements. I thought we'd go through them and see which ones we accept and which we need to discuss.
John: Could a pragma be attached anywhere?
Bethan: I tried not to answer any of those questions; we need to answer those questions.
Steven: In that case, I object to "always attach"; it "may be" attached.
Some discussion of what it means to have a pragma that is or isn't attached to something.
Nico: It needs to be attached because that establishes the scope.
Steven: If it says "use parsing XYZ" is that still attached to something?
Bethan: Yes, that's for the whole grammar.
Further discussion of what an "unattached" pragma might mean.
Steven: I think the content of a pragma might be attached, but not the pragma itself.
Some discussion of where the pragma might appear in the XML representation and what that means.
Some discussion of what the word "attach" means.
Steven: I can imagine a pragma might want to talk about some structure...
David: Might it be clearer if we used the term "scope" instead of "attach"
Bethan: That's also in the description.
John: We do know one thing, in terms of the iXML grammar, we have atoms, the symobols. We know that pragmas cannot be inside an atom, inside a terminal string for example.
Bethan: Yes, I think that's right.
Bethan: If all of the pramgas were placed at the end of the grammar and said what they annotated; I'd describe them as being attached to the grammar and expressing their scope in the pragma data.
… but in the XML, they'd have to be a child of the root element. But if we place them physically next to the constructs they annotation, we'd probably want them to be children.
Steven: I don't agree.
Nico: If you wanted to have a pragma that applied to many nonterminals, you might want to put it at the beginning of the grammar.
Further discussion of what "attach" means.
Bethan: It means it has a syntactic relationship with some element in the grammar, for example in the XML representation.
Nico: Aren't "attach" and "scope" inverse?
Fredrik: Thinking about the use of pragmas; one use I'd like is to add hints to the grammar so that the tool could know that, for example, the author thinks this is a top-down parser.
… It could be useful if you see that there's a part of a grammar that's top-down.
… And then, my attention is to annotate a rule and its children. In that case, I associate with the pragma with the rule nearest it somehow.
Steven: I see pragmas as a place where software can do stuff.
Norm: I think it's perverse to say that pragmas are "just comments" and that how they are bound to constructs is entirely up to the individual implementations.
Bethan: Where does the pragma go in the XML?
Steven: I haven't been convinced there's a reason to cast this into the specification.
Bethan: So two different iXML processors could produce different XML serializations of the same grammar.
Steven: Yes.
Bethan: That's what I'm trying to avoid.
Fredrik: I think it's obvious that the pragma should appear in a predefined place in the XML based on the grammar.
Steven: Yes.
Fredrik: Maybe the requirement formulation should say something about where the pragma will appear.
Bethan: There's nothing in the requirements that says pragmas are comments. We have to decide that.
… The terms "attach" and "scope" are just intended to help us understand what the requirements that follow mean.
… Attach is about the syntactic placement and annotate is about the semantics.
… We need some way to talk about the syntactic placement and semantic scope of pragmas.
Steven: They're only useful if they're unambiguous.
John: We have some precedent in one sense, we have comments in the definition of iXML and we have mapping of exactly where there go in the XML tree.
… It's not reversible. A comment that sat on one side or another of a ":" isn't necessarily going to go on the same side if you go back to text.
… I think we need a good mapping for pragmas.
Bethan: That's useful. Stepping back a little, the definition doesn't say anything about the placement, it just says that when we see a pragma, we know where it goes in XML.
John: Why did you decide that comments must appear in the XML?
Steven: I didn't. I originally deleted them all. Michael insisted that they be in the XML.
John: Comments can move around.
Steven: Is there a difference between "attach" and "positioned in some syntactic construct"?
Bethan: I wouldn't necessarily say that a pragma is positioned in a construct; that's less general.
… I think attached is somewhat more neutral than positioned at or in or before or after.
Bethan: Would it help to move onto the requirements and see if our concerns about the definitions actually interferes with understanding.
Bethan: Rule 1: A pragma must not change the semantics of an iXML grammar - that is, it must not cause the processor to produce a parse tree which could never be produced using the same grammar if the pragma were removed.
Steven: I'm not sure, I'd have to think about it more.
Proposal: add "according to the specification" in rule 1.
Consensus: with that amendment we accept rule 1.
Some discussion of the use of deep equal and a canonical serialization.
Bethan: Rule 2: Removing a pragma from an iXML grammar must not affect the syntactic validity of that grammar.
Consensus: accepted
Rule 3: Pragmas must not be an important part of iXML.
Norm: What does that mean?
Steven: It means they aren't necessary.
Nico: So a processor that completely ignores pragmas is still conformant?
Steven: Yes.
John: In the spirit of Michael, you could get a case where by putting in a pragma you could get answer out.
Norm: Could we just say "support for pragmas must be optional?"
David: That seems clearer.
Some discussion of whether or not even needed if it's implied by the first two rules.
Bethan: Rule 4: Pragmas must only be used for communicating with software.
Steven: I think I said "are for" not "must only".
Bethan; Yes, but as a requirement or desiderata it needs to have "must" or "should" in it.
… an existential statement is harder to evaluate.
… I'd be happier to change it to a "should".
Steven: It doesn't really express what I wanted to say.
Further discussion of whether or not "pragmas are for communicating with software" is a useful benchmark.
Steven: In that case, I'd prefer a "must".
Consensus: accepted
Some discussion of how we should have a discussion about this.
Consensus drifts towards email.
Bethan: Rule 5: Pragmas must be defined in the iXML specification without consideration of, or reference to, the needs of any particular processing software.
Consensus: accepted.
Any other business
Possible regrets from John.
Steven: And remember that the CfP for both MarkupUK and Balisage are now open.