- From: Stian Soiland-Reyes <soiland-reyes@cs.manchester.ac.uk>
- Date: Fri, 3 May 2013 09:51:50 +0100
- To: Markus Lanthaler <markus.lanthaler@gmx.net>
- Cc: public-rdf-comments@w3.org, Linked JSON <public-linked-json@w3.org>
On 3 May 2013 09:39, Markus Lanthaler <markus.lanthaler@gmx.net> wrote: >> Turtle allows the use of a relative base, as does HTML (AFAIKR), if >> @base is seen in the context of an already established base IRI, this >> could be valid, IMO. > > But neither Turtle nor HTML have external contexts. So are you really > proposing to allow relative URLs in contexts? After this discussion, I think it would only make sense to allow relative IRI references (and @base) in contexts if they are scoped to within that context. From a programming perspective it has to do with scoping rules; and we have to kind of choose if we go for PHP-style or Python style name scopes. To me (with a Python background ;) ), this would feel more natural if the scopes worked something like: A context can have a @base A context can have relative IRI references The @base of a context can be relative (to the location of the context) A context can only see names from the contexts it explicitly imports (ie. it can't see names from the previous context) @base is not inherited from an external context Relative IRI references are resolved "locally" - ie within the external context --> names imported from an external reference are always absolute This would probably also be more familiar to the semantic web world. But that would be a big jump from the current "Just keep evaluating in a single namespace" mechanism; and probably require rewrite of most parsers - it would also not allow these kind of tricks: http://bit.ly/16tFZql { "@context": [ { "id": "widget://c13c6f30-ce25-11e0-9572-0800200c9a66/" }, { "proxy": "id:.ro/annotations/" }], "proxy": "bananas" } -- Stian Soiland-Reyes, myGrid team School of Computer Science The University of Manchester http://soiland-reyes.com/stian/work/
Received on Friday, 3 May 2013 08:52:37 UTC