Re: CURIE objections in HTML5+RDFa

[CC trimmed per previous instruction.]
On Feb 18, 2009, at 18:00, Manu Sporny wrote:

> Henri Sivonen wrote:
>> Using full URIs exposes the cost of the actual usability problem
>> to the RDF community instead of making adjacent communities bear
>> the cost through complication to their formats.
>
> Let me attempt to summarize your e-mail.
>
> You believe that full URIs are less bad than CURIEs, but feel that  
> full
> URIs are still a bad solution for the semantic web. You are concerned
> that if RDFa fails to garner decent adoption rates, that HTML5 will  
> then
> contain unnecessary cruft for supporting CURIEs.

Also, if RDFa turned out to be successful in text/html (with or  
without a blessing by the HTML 5 spec), we'd be left with syntactic  
complexity in the platform. In particular, if RDFa succeeds for a  
couple of use cases and fails in general (or succeeds otherwise in far  
down the long tail of use cases), the important use cases would be  
stuck with complex syntax.

I'm particularly worried about ccREL succeeding to the point that an  
alternative solution can no longer be launched into the market to  
replace it and Free Culture then getting encumbered by the syntactic  
complexity preventing even further success.

> Could you provide at least one alternate mechanism? The mechanism  
> should
> not use full URIs, and should addresses most, if not all, of the
> problems solved by using full URIs?

A backwards-incompatible alternative mechanism would be tokens of the  
type "prefix-local" (or "prefix:local", but I'm trying to avoid  
confusion here) where prefix *wouldn't map to anything*. That is,  
processing would merely compare the "prefix-local" code point for code  
point without expanding it to anything. Prefixes would be from two to  
four letters--preferably acronyms for the vocabularies--allocated on a  
first-come-first-served basis either through a central registry or  
through best-effort collision avoidance by the community.

Now, how to make this backwards compatible with the RDF model? If the  
predicates don't need to be dereferencable, a one-letter URI scheme  
(e.g. 'r' for RDF) could be registered adding two characters of  
overhead per predicate: "r:prefix-local".

To add back dereferencability in pre-existing software and to use a  
pre-existing registry system, a TLD called 'rdf' could be registered  
and the identifiers could take the form "http://local.prefix.rdf" with  
11 characters of overhead. If a software update for dereferencability  
is OK, "r:prefix-local" could be defined as the identifier to compare,  
but to dereference it you'd map it to "http://local.prefix.rdf" before  
passing it to the HTTP layer.

-- 
Henri Sivonen
hsivonen@iki.fi
http://hsivonen.iki.fi/

Received on Thursday, 19 February 2009 08:32:17 UTC