[topicmapmail] PSIs - alternatives

[Alexander Johannesen]

Hiya,

On 6/20/06, Steve Pepper <pepper at ontopia.net> wrote:
> I wouldn't say that the subject has been discussed ad nauseam.

Hmm, every camp has had long discussions on this very topic for years.
The point wasn't that we shouldn't continue to talk about it, though,
so please, do carry on. :)

The whole "difference between topic mappers and semWebbers" also seems
to culiminate in the PI debate. There's bucketloads of different
attempts to "solve" this, partly or in whole, from Purple Links to
URN's to Furl to all sorts of varieties. It's a problem within www
tagging as well as semweb where URL's are identifiers, it's a problem
within blogging (where bloggers want to identify their discussion
partners, mostly resolving to FOAF'ing), it's a problem in knowledge
systems, it's a problem because people can't tell identifiers from
names, and it has been discussed around and around. Just because we
have no solution doesn't mean it hasn't been talked about a lot.

> In fact, I
> think Simon raises very important points. The PSI mechanism is fundamentally
> sound but it is not being used as widely as it should, mostly because of the
> lack of a discovery mechanism.

I remember when I was but a wee lad on this very mailing-list, raising
the exact issues. :) There were no drive towards centralised resolvers
then, I haven't seen any drive towards that of late, some think it's
the wrong idea (centralised anything being bad), while others are
waiting for someone else to just go ahead and do it.

As to PSI's being sound; well, if it *was* sound, then we would all
use it. Perhaps it's a marketing problem, though, but to me the notion
that the PSI itself is really without meaning until you attach some to
it I think is hard for a lot of people to grok. I only see aggregated
and federated PSI's as the way to truly solve this (who knows; a
metamodel of PSI resolving?), but that would be a centralised resolver
service, and who's willing to host, service and maintain such a thing?
How do you do this decentralised? SemWeb people tends to solve this by
... uh, not solving it, so what are our options?

...

> > And since your PSI doesn't actually have to
> > point to anything that exsists, it doesn't matter that the PSI as a
> > URI works, and that means that people not only can but *will* break
> > them over time.
>
> Actually, to be considered a PSI (i.e., a Published Subject Identifier),
> according to the OASIS recommendations, a PSI must resolve to a subject
> indicator. That's the whole point. Why publish an identifier if there's no
> way for potential users to figure out what it is supposed to identify?

And you have never used a PSI that didn't resolve to something? :) I
think there is a difference between reality and practice here; we
would *love* our PSI's to resolve to something we could use straight
away, but at most you get HTML page with some text on it which isn't
useful for machine parsing. So whether the PSI is real or imaginary
doesn't really matter as long as the meaning of it still relies on
human interaction and for this human to model it in their respective
topic maps.

> As I said, I think the PSI approach is fundamentally sound and I have been
> pushing it in the W3C recently, albeit under a name that relates more
> immediately to what people in the W3C know and understand. [1] The response
> so far has been very positive and I am hoping to initiate an activity that
> involves W3C, SC34 and OASIS in the near future.

And this is good progress indeed; glad to hear of it.


Alex
-- 
"Ultimately, all things are known because you want to believe you know."
                                                         - Frank Herbert
__ http://shelter.nu/ __________________________________________________