Re: Banging the identity drum again, this time REST-style

Lars,

Just quickly as I think Andrew managed a pretty good response. At least
it wasn't profane. ;-)

I really tire of topic maps being described as having "...lost the
game." 

I really don't know what game we were supposed to have won? 

The latest I heard from the Linked Data crowd is that subject identity
is caveat emptor. I suppose that applies to URIs.

And the number of XML formats, database schemas, OWL ontologies, etc.,
hasn't slowed any that I have noticed. 

Granting that topic maps aren't as popular as I would like, if anything,
the market for the application of topic maps has simply gotten larger.

Case in point. I blogged today about a new approach to compressing
patterns in large data sets. Very exciting.

OK, so how am I going to merge my results of compressing such patterns
with yours from six months ago? Assume someone just handed it to me.
Doesn't look all that odd. ;-) 

I could guess at what you meant, convert your data to mine or mine to
yours. Not really all that satisfactory. Because I am *guessing.*

It's the indexing problem all over again. 

OK, time to 'fess up. It is possible that any legend I write to treat
the two data sets as the same, may make the same mistake. But hopefully,
if I am working from a legend that you wrote and the one that I wrote,
the odds will be somewhat lower. 

NB: lower, not non-existent, lower.

If it makes you feel any better, take note that the SW has spent
literally $billions only to say you have to determine what subjects URIs
represent for yourself. I take that as fertile ground for topic maps. 

Hope you are having a great evening!

Patrick


On Mon, 2010-11-22 at 23:59 +0100, Lars Heuer wrote:
> Hi Andrew,
> 
> [...]
> > I'll have to actually READ the RFC
> 
> Progress ;)
> 
> > and have a think to respond
> > properly, but I think you inverted the problem.  I was talking about
> > what Topic Maps could bring to the RFC (based on reading the intro
> > and assuming that it is the basis of the link I referenced inside the snark tag).
> 
> Call me pessimistic, but we *have* a Link RFC. You may know that, but
> an RFC is a standard. You may call the editor and mention that this
> and that is suboptimal. You can may even call me (if you'd know my
> mobile cell no) and tell me that CTM sucks and I'd tell you "The train
> has left the station" (maybe I'd ask some questions about Bob the
> Builder, though ;) ).
> 
> > I think we need to help the world understand how Topic Maps can
> > help in these situations rather than implementing
> > yet-another-directory or yet-another-identity-mechanism.  If you
> > need one, fine, but at least be able to express it in terms of
> > things that exist so people can understand why it's relevant and necessary.
> 
> Maybe Topic Maps (TMRM more than TMDM) is necessary, but we don't have
> a big PR department. Topic Maps never had and we lost the game, I'd
> think. We can offer mappings, though. Topic Maps isn't bad and I think
> mappings to Topic Maps offer some possibilities aside from portals.
> 
> RDF/LOD hasn't solved the problems of subject identifier vs. subject
> locator. And the current players of RDF/LOD don't understand that
> problem properly or deliver poor solutions. But who cares? I care and
> few people may care, but the current trend goes to "something
> has-relationship-to something-else". Take Atom, take ActivityStreams,
> take anything which is popular, it's all about triples. IRI
> reltionship-IRI IRI. That's enough. Well, you have to take care about
> statements like
> 
>    john father-of sean.
>    sean child-of john.
> 
> but who cares? It works.
> 
> Well, now I am far away of your "Link" header proposal and I should
> read your response to Patrick. :)
> 
> That's the same discussion as "Prolog vs. <ANY-LANGUAGE-HERE>".
> Java/Python/Ruby is popular. Prolog is cool, but who cares? ;)
> 
> Best regards,
> Lars