Re: Real data

> Sandro Hawke wrote:
> > So where are you thinking you'll serve the RDF?
> > 
> >
> > 
> > would be obvious, but including the ".rdf" in the URI for the protien
> > itself doesn't seem right.
> The default representation if no extension is specified is a normal web 
> page. I guess you could try to do something with content negotiation, 
> but I'd rather not rely on that. On the other hand the page could 
> include something like:
>    <link
>      rel="alternate" type="application/rdf+xml"
>      href=""
>    />

Right.  I think that's probably the best bet right now, but other
people may have their suggestions.  I'm not sure what work the "Best
Practices" working group has completed on this issue. 

> Note that if you actually entered, 
> you would be forwarded to some other site that is capable of serving the 
> resource. 

I understand; that seems fine.   Any serious RDF dereferencing module
should (IMHO) be following redirects, as well as link/alternate
references as above.

> Since these URLs are not really as stable as they should be, I 
> prefer to use URNs for identifying resources, e.g.

I totally understand the temptation.  I'm still co-author of the tag:
URI spec [0], which addresses the problem that "lsid" isn't actually a
registered URN namespace [1] [2].

But I strongly encourage you to resist this temptation!  It's so much
more cool to be able to actually surf the data, using the URI/name for
a protein/person/property/whatever to automatically and efficiently
get more data about it.

> This also avoids confusion with users who always expect a URL to be 
> resolvable...

But there's no problem, as long as you make your URLs resolvable,
right?  And you probably want to do that, since each one is an
opportunity to be useful.

      -- sandro


Received on Wednesday, 21 July 2004 15:24:19 UTC