Serialization

2002-05-17#3  jjc  Propose new text to describe serialization of b-nodes.

Sorry I had forgotten this.

Proposal:

Delete section almost entirely, and just document what cannot be done.

Rationale: we only owe the world an explanation of how to do what is 
unreasonably hard. There is published literature about RDF serialization, 
there is no need for the REC to duplicate this unless it is *necessary* for 
correct implementation.

e.g. (sketch)
[[
Serialization
==========

There are certain  RDF graphs which follow the abstract syntax that cannot be 
serialized in RDF/XML.

These are those that:
- use property names that do not 
end in a (possibly empty) sequence of NC_NAME characters precededed by an 
NC_NAMESTART character preceeded by a URI ref. i.e. those that do not match
the regular expression:
  URIref NC_NAMESTART (NC_NAME *)

or
- use inappropriate reserved names as properties.

]]

Have I listed all the problem cases?

If we wanted to reference my work on serialization, a short para like:

[[
Carroll [ref] discusses techniques to serialize RDF/XML in a fashion that uses 
the full range of grammar rules to create human readable output.
]]

Neutrally, I don't believe that that para is necessary, nor do I think it 
would be inappropriate.

Jeremy

Received on Thursday, 24 October 2002 23:35:47 UTC