Re: bad language tags

On 08/05/13 09:29, Antoine Zimmermann wrote:
> Le 07/05/2013 13:25, Andy Seaborne a écrit :
>>
>>
>> On 07/05/13 10:37, Antoine Zimmermann wrote:
>>> Well, what does the N-triples spec says? I would like it to say that
>>> "xyz"@en and "xyz"@EN both correspond to the language-tagged string that
>>> has "en" as its language tag.
>>
>> I find "correspond" ambiguous.  Something can correspond to several
>> things.
>
> So, here is how I see things: a data model is an abstraction that is
> formalised mostly in terms of mathematical structure, often in set theory.
>
> A serialisation format for a data model is two things:
>
>   1. A grammar that discriminates valid and invalid sequences of
> characters for that format;
>   2. A function from the valid sequences of characters to the
> mathematical structures of the data model.
>
> Applied to RDF and N-triples, what I said is the following:
>
> """
> The function that maps a string of characters (conforming to the
> N-triples grammar) to an RDF graph, maps "xyz"@en and "xyz"@EN to the
> same language-tagged literal that has "en" as its language tag.
> """
>
> Or, if we want to go along this line, "xyz"@EN is not a valid N-triples
> representation of a language-tagged string (i.e., it is not conforming
> to the grammar).

I prefer EricP's characterization - language tags are values in the 
abstract data model.  None of this putting lexical representation of a 
language tag into the abstract data model.

You can write a value down in different ways.

>> N-Triples (RDF test cases) requires lower case only.
>
> Fine. It matches concepts more directly. And if Turtle allows upper
> case, it is fine too.
>
>
>> N-Triples (in draft - motivated by being a dump format) follows Turtle.
>
> Not sure what you mean by "follows" here.

Has the same grammar rule.

	Andy

>
>
>
> AZ
>
>>
>>      Andy
>>
>>
>
>

Received on Wednesday, 8 May 2013 11:05:18 UTC