RE: Non-Text Literals

> From: Ashley Yakeley [mailto:ashley@semantic.org] 
> 
> At 2002-08-31 05:14, Bill de hÓra wrote:
> 
> >Don't inline the photo as a literal.  Give the photograph a URL and 
> >associate it with you:
> 
> That's not a solution to the problem. I want to hand someone 
> something 
> self-contained, I don't necessarily have a useful place on 
> the web to put 
> things, nor may the reader have internet access at the point at which 
> they wish to examine my document.

Ok. Self contained in RDF is handing someone an RDF graph. Sending
resource representations around with the graph is extra.  I would have
thought you could publish the RDF without the inlined image and said
things about the image resource by giving it a URI/L. You don't
necessarily need to be online to do that and I don't necessarily need to
be online to understand your RDF. Certainly I don't need a download of
the image's representation to understand your RDF.

I'm not saying inlining is bad, but it has its own set of issues; like
any optimization, it may become tomorrow's bottleneck. RDF evangelists
will tell the big win is in /reusing/ RDF information. IMO the way to do
that is give things URIs. The interesting things happen when you make
your image a resource and you, I, and anyone else can share RDF
statements about it. There's only so much you can do with literals since
they can't be subjects of statements. For example. today there's no
standard way for I and someone else to know (infer) that the literal
image we have is indeed the same image you sent out twice. Ultimately,
BLOBs need URIs to make them truly valuable in RDF.

How do you and the readers share information? 

> Of course this may be exceeding the expectations for RDF...

Examining images is outside RDF, you can only reason about the graph.
And at some point you'll need a way to share the RDF+literal image, even
if that's handing someone a floppy or sending an email. Why not hand the
URI over instead?


Bill de hÓra 
--
Propylon
www.propylon.com

Received on Sunday, 1 September 2002 14:17:13 UTC