Re: I18n and Linked Data - an important (but fixable) omission?

Hello Tom, others,

On 2011/09/10 3:21, Tom Baker wrote:
> On Fri, Sep 09, 2011 at 06:08:21PM +0100, Jodi Schneider wrote:
>> This complicates the section on Linked Data -- one of the key places I think
>> we need to simplify. So I would propose reverting that change, so that this
>> paragraph focuses only on Linked Data -- the concept it is defining.
>
> I take your point that squeezing a reference to IRIs into a definition of
> "Linked Data" interrupts the flow of that brief definition.
>
>> Then, if we do feel the need to cover URIs in the Scope section, I'd rather
>> that we gave it its own line (similar to how we define "libraries").
>
> That sounds like a good solution.  I think the section should end with
> "Library Linked Data", so my preference would be to insert a new item
> between "Linked Data" and "Open Data" -- i.e., right after "Linked Data".

That sounds like a good idea.

> Maybe it could be called "Uniform Resource Identifiers (URIs)", define
> URIs, and refer also to IRIs.

It's unfortunately somewhat hidden, so you may not be aware of, but in 
terms of technology, in particular RDF, what it calls "URIs" are 
actually IRIs. Please have a look at 
http://www.w3.org/TR/rdf-concepts/#section-Graph-URIref.

So (very roughly paraphrased), to write something like "Linked Data uses 
URIs (ASCII only), but you might also think about supporting IRIs", 
would be wrong. It would be better to write something along the lines 
""Linked Data uses URIs. By definition, this includes IRIs (see Section 
6.4 of RDF Concepts)."


>> Alternately we might want to put it in the "Available Technologies" Appendix
>> section of the report: We have considerably simplified a number of issues in
>> the main report.
>
> I think IRIs are important enough to emphasize up-front -- in the Appendix,
> the point would be much less prominent.
>
>> While I'm not sure that the *term* "IRI" is that much harder to understand
>> than "URI" (which is different from the "URL" which is in common practice),
>> you make a good point that URIs, rather than IRIs, are currently emphasized
>> in Linked Data. It would be helpful to know whether, for instance, the
>> National Diet Library is currently using IRIs for Linked Data.
>
> ...or indeed, whether the advocates of IRIs advocate their use in libraries
> regardless of scripts used -- i.e., even for Latin-script URIs?

I'm not sure I understand your question here.

If you want to ask whether libraries e.g. in the US should make sure 
that the semantic web technology and products they use conforms to the 
specifications and does not limit identifiers to US-ASCII only, then the 
answer would be clearly YES.

If your question is whether a library e.g. in the US should use an URI 
or an IRI for an identifier such as 
http://en.wikipedia.org/wiki/Football, the answer is that this is 
irrelevant; by definition, all URIs are also IRIs.

If your question is whether such a library, for such an identifier, 
should add non-ASCII characters to make it an IRI but not an URI, e.g. 
by modifying the above URI/IRI to something like 
http://en.wikipedia.org/wiki/Fóòtbåll, then the answer is of course NO. 
I hope nobody advocates such nonsense.

If your question is whether a library e.g. in Germany or France or 
Italy, where the languages used are written with the Latin script 
including diacritics, should create IRIs that are not URIs as 
identifiers, then this may depend on various circumstances, i.e. the 
availability and familiarity of people with US-ASCII fallbacks,... As an 
example, the German Wikipedia has http://de.wikipedia.org/wiki/Fußball, 
but this is also available under http://de.wikipedia.org/wiki/Fussball.
(Please note that I wrote 'create'; for 'use', the answer is different 
because identifiers may come from the outside without a choice.)

If your question was about something else, please tell me.

Regards,   Martin.

> Bottom line:
> since this report is about Linked Data, and the Linked Data message always
> talks about URIs (or even URLs), I think we need to stick with URIs.  But we
> can and should draw attention to IRIs up-front.  Inserting a separate item
> into Scope would do that even better than the solution I proposed.
>
> Would you like to propose a text?
>
> Tom
>

Received on Saturday, 10 September 2011 03:07:49 UTC