Identity problems numbers 3 and 5 - was Re: Toward easier RDF: a proposal

(Subtitle: Is SQL really easier to represent knowledge than the Semantic Web?)

I think I said pretty much what I had to say on the URI generation and related addresses etc stuff, probably more than once.

I leave the discussion feeling rather disappointed.
You can consider this a rant if you like, but it is not meant to be.

I don't think that I have got any bits of code or suggestions that will actually help me or other developers to actually do practical stuff.
Nor really some more strategic ideas to get SemWeb more usable, as David hoped for.

On the contrary, I have been left with the feeling it is far too hard to do addresses, or even anything in Semantic Web technologies - maybe SQL would be easier?
https://www.google.com/search?q=representing+addresses+in+sql
is certainly pretty helpful.

I know this was not the intention of the discussion participants, but that is certainly a reasonable view to take away.
Maybe this is the wrong forum - is there somewhere else where I would get help?

What do I conclude from this?
You cannot have a discussion about how to do something on the SemWeb without vanishing down the rabbit hole of what is identity.
And then the discussion almost becomes:
"Why would you want to do that?"
"You can't do that! Because..."
"That's too hard - let's think about an even harder problem!"
Meanwhile, over on StackExchange, example SQL tables will have been created for me.

Could the answer on address representation not be:
"Great, we have this vCard schema (first hit on Google, by the way). Use that carefully and you will probably be alright."
And even better if there was a cloud of experience, examples and code around it that helped me use it and interwork with other tools.

I do know that identity is hard (dog knows I have suffered that).
My favourite phrase in this is attributed to Cratylus: "You cannot step into the same river once".
So I do deeply share peoples' worries about identifying things.

But these problems are absolutely standard and common across all the domains we try to model.
I am reminded of the joke about comedians sitting in a room and just saying numbers and then laughing - (each number is a different joke that they all know).
So Identity Problem number 3? - "The world is infinitely complex, so you won't be able to accurately create a URI for anything."
Identity Problem number 5? - "The world changes, so none of your URIs will be stable over time."
Identity Problem number 4? - "Things have multiple identifiers - you won't know which to choose."
Identity Problem number 2? - "Yeah, but things don't look the same in other cultures."
Then there are others which crop up occasionally, but aren't actually problems, such as "But people will be able to make incorrect statements about my URIs".

These problems are not unique to SemWeb - they are common to any development anyone does.
They are nothing to do with the technology chosen - their existence may be clearer in SemWeb, and that is a Good Thing; and it may also mean that we can address them better in SemWeb, or at least understand what the implications are.

So if someone asks for advice on representing dogs, say, could the answer not be:
"Oh, dogs suffer from Identity Problems 1, 2, 3, 4 and 5." (if we must give the warning)
(I suspect that everything suffers from all the problems, and many philosophers have listed them.)
And then just move on.
"Try http://www.wikidata.org/entity/Q21078, which will give you these: https://commons.wikimedia.org/w/index.php?title=List_of_dog_breeds"

Best
Hugh

By the way, here are some fo the routines I use, if that is of interest to anyone:
https://docs.google.com/document/d/1HjjypKzMWGUuVd818W67ODVXU3W0fhpef9qYX0PQhO8/edit?usp=sharing


-- 
Hugh
023 8061 5652

Received on Friday, 7 December 2018 14:08:42 UTC