Re: SYNTAX: RDF/XML Syntax WD work

Hi Dave 

With the danger of rushing in before you've reached anything like a
stable point ...

I have had a look at version 1.88 and I am decidedly nervous about the
direction you're going with the triple production stuff.

I fear you're heading for a rerun of M&S section 6 with:
+ too many words
+ that are self-contradictory 
and 
+ too unclear


I am also less than taken with the view that the text describing the
production should be associated with a particular processing model. 

For instance in the subsection "Production node" you say 

[[[
       The following processing is performed before handling any
propertyEltList children and in this order.
]]]

That is presumably intended in an abstract processing model. 
It is in my view, perfectly legal to process the property attribtues
after the property elements, and the resulting n-triple file is an
acceptable rendition of the RDF/XML.

I would go further and say that the property attributes and property
elements can be processed in any order with the single exception that
rdf:li is converted into rdf:_NNN in document order.

I also dislike the "in this order" part since many of the steps you then
list are, in fact, interchangeable.

I fear we have a stylistic clash brewing.

I feel that it is better if a specification does not over-specify by
indicating superfluous things; whereas you appear to want a
specification that would have helped you write the particular
implementation of RDF/XML that you have in mind. 

I do not feel that it is the job of a specification to assist a parser
writer. The job of a specification is to specify what needs specifying
(and no more or less). If at any point there is a conflict between the
task of assisting the reader in their task (e.g. writing a parser), and
the principle aim of specifying I am clear that the reader should be
left to seek assistance elsewhere. The reader should expect clarity from
a spec, if they want a howto then they may need a different document.

Jeremy

Received on Thursday, 18 October 2001 04:50:45 UTC