Regrets for tomorrow's teleconf. and summary on DTB discussions.

Sorry, this might be unconvenient, but due to an urgent meeting whch I 
cannot shift, I have to pass on tomorrow's telephone conference.

So, in order not to hamper progress for DTB, I suggested several options 
to vote over concerning the points in my mail at [1].

1) CURIEs: As for 1) there were extensive discussions on the 
mailinglists, it seems that [2] is kind of a minimalistic proposal 
whereas [3] is richer. I haven't seen a grammar for [2] yet, so let me 
put both options on the table again.

a), cf. [3]

   ANGLEBRACKIRI     ::= '<' IRIRef '>'


   CURIE                ::=  PNAME_LN | PNAME_NS

   Const             ::= ANGLEBRACKIRI
                       | CURIE
                       | STRING^^ANGLEBRACKIRI
                       | STRING^^CURIE

b), cf. [2]

   ANGLEBRACKIRI     ::= '<' IRIRef '>'


   CURIE                ::=  PNAME_LN | PNAME_NS

   Const             ::=  CURIE
                       |  STRING^^ANGLEBRACKIRI
                       |  STRING^^CURIE

Comparison between a) and b): The only difference is that b) doesn't 
allow ANGLEBRACKIRIs as Consts, thus making it N3 incompatible, but 
well. Both are context-sensitive. There were some other discussions 
introducing some form of "aliasing" [2,4], but since I didn't see a 
grammar for this and thus it is unclear whether these would introduce 
ambiguity,  I suggest to keep it out.

I suggest to vote between these two, my own vote is for a), though I am 
willing to obey a majority vote for b). I personally would be unhappy 
with N3 incompatibility [5,6] when voted for b), since none of the 
arguments given so far were technical in the sense of that there would 
be any problem with the grammar for a). For an additional argument, see 
also 4) below.

2) FULL URIs for RIF (see also [7])

 From the original proposals, the following 2 seem to have "survived" 
the discussions so far:

a) define own prefixes (separate for functions and predicates):

c) reuse XPath/Xquery fn: prefix (problem: not prefix defined for op: we 
still would need to invent one):

I personally prefer a) and suggest to

PROPOSE: We define own namespace prefixes
   PREFIX("pred", "").
   PREFIX("func", "").
for RIF builtin functions and predicates

3) The handling of errors seems not to be anything we need to discuss 
over again.

4) Additionally, I suggest to introduce:

   PREFIXDef         ::= ' PREFIX(' PNAME_LN , STRING ') .'

for the prefix definition in the presentation syntax.
Whether this expands to Qnames or entitties in the actual XMLificaiton 
is a separate issues [8,9] and not important for stabilitzing DTB, it 
seems. In doubt, I am with Michael here [9] and suggest that in a 
translator to XML PREFIXDef translates to an ENTITY definition... but 
that's an implementation detail anyways, one could likewise simply 
expand all CURIEs in the XML.... The only problem with that is that if 
you want to translate *BACK* to presentation syntax again, you will end 
up with something ugly, if we go for option b) on 1) above.









     and following thread.


Received on Monday, 5 May 2008 11:48:40 UTC