danc-02 (was Re: open concepts issues)

>>I have made proposals on
>>danc-02 goofy literals
>>however they are now out of date given our literals decision.
>>
>>I will make a new proposal, but probably not in time for the telecon

>I would very much like to have a proposal for this weeks telecon.  Is there 
>some way to help get this done?

with graham's corrections, the only thing that is missing is actual text.

I hope to provide some this afternoon or tomorrow morning (very early tomorrow 
morning for the US people !)

The problem is:
1: minimizing the risk of misreading concepts to give
<eg:a> <eg:p> <eg:c> .
and
<eg:a> <eg:p> "eg:c".
the same abstract representation.

2:ensureing that the  abstract representation of:
<eg:a> <eg:p> "00"^^<eg:en> .
and
<eg:a> <eg:p> "00"@en
differ by more than the three characters "eg:"
(It is fortunate that the set of absolute URI references and the set of 
language tags are disjoint, but I am not happy to have an abstract 
representation that relies on that).

3: not introducing new text/framework etc to make these distinctions
(don't be too heavy - objectives 1 and 2 are obvious really)

4: not being too goofy

I felt that the old text, with the old literals (including all three 
components) adequately met these objectives (while Dan differed on 4). But it 
is quite fragile and I don't want to bodge it.

In my head at the moment is something like:

[[
A literal in the RDF abstract syntax is either a plain literal or a typed 
literal.

A literal has one or two ?named? components.

All literals have a *lexical form*. (Defn of lexical form from current draft)

A literal with only the lexical form is a plain literal.

A plain literal may also have a language identifier (defn of language 
identifier from current draft)

Typed literals have a "datatype URI" (defn of dattaype URI from current draft)

]]

The earlier informative text should be much easier to read e.g.
In RDF there are:
- plain literals consisting only of a string
- plain literals consisting of a string and langauge identifier
- typed literal consisting of a string and a datatype URI


Polish is needed - working from the WDs and any relevant mail traffic, and 
looking at any changes made by Graham in the editors draft.
The sketched normative text maintains the style that DanC-02 objects to, but I 
find it the simplest way of meeting the objectives 1-4 above, (well 1-3 at 
least); and I am hoping that the changes in the semantics address the 
underlying issue.

Jeremy



.

Received on Thursday, 15 May 2003 04:23:20 UTC