The Mr. Peppers Minting PSIs

CatBert was in a very aggressive mood. He had interviewed me all day on subject identification and how it worked for the Topic Maps and the RDF stack.

But the longer he investigated, the more poignant his remarks became. This evening he even meant:

http://kill.devc.at/system/files/angry-cat-small.jpg Robert, I am getting tired of the lack of imagination in the semantic web community. That applies to both, the Topic ShMaps folks and the RDFreaks.

Identity Issue

I had told CatBert, that - when attending RDF related conferences - then the semwappler'sh accolytes are completely ignorant about RDF's weakness to properly address subjects. Not the SemWeb superstars, though, but they choose to ignore it as they want to go down to business with their tuple stores.

Inspite that, the discussion

  • What does an IRI actually mean?

would continue to crop up every now and then. But it is too late to change anything. The httpRange-14 train has left the station, slowly rolling down the hill.

I had also told CatBert that the Topic Maps people are more alert to the issue, and that subject identifiers and subject locators have a better - albeit not perfect - approach to this problem.

But CatBert was very unhappy.

A World of Stakes

http://kill.devc.at/system/files/angry-cat-small.jpg What I am getting tired of is the IRI fetish with which people try to tackle this: Every subject must get an IRI. Putting stakes into the problem. As if this were some sort of vampire movie...

He's right. The obvious problem with that is: Which IRI(s) to (s)take?

Many people have discussed before whether Wikipedia should be the "Gold Standard", easily refuted by those who observe that WP's view is not an absolute one but has cultural context. Like everything else.
Others propose meta-identifier repositories noisily moving the grounding problem around without actually achieving anything.

CatBert became cynical:

http://kill.devc.at/system/files/angry-cat-small.jpg And is this where you stand? What comes next? Another committee, perhaps?

I managed to produce a forced smile.

Tip Of The Vampire

And sighed. The real depressive message is that this is only the tip of the problem iceberg as subjects can be viewed in different resolutions.

The canonical example is the city Berlin: On the one hand it is a geographical thing. But there is the political aspect, the administrative one, and also the historical. And many others.

None of these issues (temporal and semantic resolution) are honored by chasing vampires.

Saving Humanity

With the courage of despair, I made a proposal:

  • What if we stop this IRI madness, and use a query language for that?

CatBert fell silent. This part of the counter attack had worked. I got his attention.

  • We could use a mix of fixpoints (IRIs) and path expressions to address what we mean. Something like this to address me personally:

"http://kill.devc.at" \ blog

It would take the literal URL and find all topics which have this as blog. And that would be me.

  • And that would address you, my dear CatBert:

"http://kill.devc.at" \ blog <-> owns
      == "CatBert" \ name

In the first line I would get all the things I own, and then intersect it with the things which call themselves CatBert.

  • See how few IRIs I had to use, and still get to you?
  • And if the language happens to understand time, then you could also write:

"http://kill.devc.at" \ blog <-> owns
       == "CatBert" \ name
       == . isa Cat after 2003

Just to make sure that I mean really you and not some other ghola incarnation. I knew that CatBert has read the God Emperor.

Ending up on the Sixth Circle

He played with his whiskers.

http://kill.devc.at/system/files/angry-cat-small.jpg Pretty cool, I have to admit. But there is no way the people will follow you there. They rather prefer to continue in their vampire mass hysteria.

CatBert was right. I will keep this to myself.

Posted In

Oh dear

You used those Dune references to trick me in here, didn't you?

There are a few suggestions floating around to allow sub-selection as a means of identity in some MIME types, so that you could mix IRIs and "some query definition" to do this through IRIs, for example http://bingo.com/animals#on_the_ark//near:the_fall//clean[1] (first animal that was on the ark and around the time of the fall, and is a clean animal, all of which is false, of course) which could lead to interesting things, however since anything after the anchor isn't used (or shouldn't be used) by servers, it won't *resolve* identity, only work on the client side. Not sure if that's significant.

I've toyed around this as well, but the problem is that it doesn't always work, especially where queries return more than one, or none.

I guess some approximation rules could be defined. The trouble with binary logic is that it is binary, and based on logic.

Alex (not verified) | Mon, 12/07/2009 - 00:25

Re: ... it doesn't always

... it doesn't always work, especially where queries return more than one, or none.

What is exactly the problem with that? It is what real life gives. Is the goal to ignore reality?

The trouble with binary logic is that it is binary, and based on logic.

Do not underestimate the degree of fuzziness in this: You have geographical distance and you have time.

Where there is quantity, there is fuzziness.

rho | Mon, 12/07/2009 - 11:54

The problem?

The problem with that is of course that if you need to attach some property to John Smith who's wife's name is Brenda and have two children together and you get two results, which one do you attach it to?

As to time, I can agree time has some merit, but I don't agree with geographical space, at least not on its own. We all move about far too often to make any serious sense (I've moved internationally 5 times in the last 5 years) of having the property attached and continuously updating myself. Isn't that the real problem, though? That my own homepage still state I've got two kids, and I'm 34 years old?

Alex (not verified) | Mon, 12/07/2009 - 12:25

Identfiers

The more I look into the problem of identifier arbitrariness the more gloomy it looks. This weekend I was thinking that perhaps identifiers should have scopes to cope with culture, resolution, etc. , but I think that this also will just push the problem further forward. Perhaps, like you suggested, some kind of identifier query language is the way to go forward, but is there no way of doing so without any fixpoints?

Inge Henriksen (not verified) | Mon, 12/07/2009 - 10:50

Re: Identfiers

... but is there no way of doing so without any fixpoints?

CatBert answers:

  • While you humans always have a tendency for absolutism (think email addresses, religion, ...), there is actually much room for relative addressing:

"The spouse": Given a person, this relative address gives the wife or husband of a person.

"The man formerly known as The King".

So - theoretically - life would work without any absolute position. But it is not practical for humans. As a cat, I have absolutely (pun intended) no problem with that.

rho | Mon, 12/07/2009 - 11:51

Fixpoints

Fixpoints themselves are just shortcuts for something that itself is a query (if going the full mile), making the problem quite recursive. The thing is that the problem isn't any more a problem in the virtual space as it is in real life, it's just that most people are law-abiding citizens shielded from some serious epistemological conundrums.

Alex (not verified) | Mon, 12/07/2009 - 12:20

Epistemological conundrums

I don't believe that the whole "what is knowledge"-discussion really has to be agreed upon to solve these identifer problems (personally I believe that any sentient being cannot know anything more than of its existance, all other "knowledge" are conjectures imho :), but that is another matter ).

Inge Henriksen (not verified) | Mon, 12/07/2009 - 13:44

Conjectures

Yeah, science is a hoax. :)

Alex (not verified) | Mon, 12/07/2009 - 22:03

Re: Conjectures

Yeah, science is a hoax. :)

You're just a mean man!

I think I like you ;-)

rho | Tue, 12/08/2009 - 08:30

Internalism

Internalism is very much a part of science.

Inge Henriksen (not verified) | Tue, 12/29/2009 - 13:20

Do we kill all the identifiers? or just pare them back a bit?

Nice essay.

Is CatBert opposed to any use of IRIs for subjects? Or only to the proposition that EVERY subject must have an IRI?

C. M. Sperberg-McQueen (not verified) | Wed, 12/09/2009 - 02:03

Re: or just pare them back

CatBert seems to be against all silver bullet approaches (here comes the vampirism metaphor again). He says, that IRIs per se are fine, but the world as pincushion? No.

rho | Wed, 12/09/2009 - 10:50