Re: ISSUE-13: History of rdf:XMLLiteral

On 10 Nov 2011, at 23:00, Jeremy Carroll wrote:
>> Well then let's make that explicit.
> 
> That basically takes us back to something like the first LC design that was rejected.

Right. My impression was that it was rejected in 2003 for two reasons:

1) OWL peeps didn't want to ship XML parser with their reasoners

2) In the DOM you don't know whether the original XML had single or double quotes, so an RDF/XML parser implemented on top of a DOM parser may not be able to exactly reproduce the XML from the input file; so why not canonicalize right in the parser to make it deterministic

I think 1) is no problem if rdf:XMLLiteral is optional in the datatype map. (That probably wasn't an option in 2003 due to i18n pressure, but we now know that rdf:XMLLiteral failed to deliver its i18n promise.)

I think 2) is no longer relevant as RDF/XML is no longer the only game in town.

> Personally I preferred putting the canonicalization in the L2V map.
> It is quite a pain to specify though, and people didn't like the verbal gymnastics required.
> (You do end up needing a wrapper element ... yuk yuk yuk, and for backward compatibility we will need to add the wrapper to the lexical form, apply C14N, then remove the wrapper ...)

I thought the wrapper element was needed because XML literals used to have language tags at some point. I'm not proposing to bring that back! I don't think a wrapper element is necessary for what we're talking about here. (But yeah it would end up being more wordy.)

> A different way of making explicit would be to explicitly state that c14n is only required where the app requires comparison.

I find that a bit messy. What exactly does that mean? How does the parser at one end of the chain know what the app on the other end requires?

Best,
Richard

Received on Thursday, 10 November 2011 23:48:24 UTC