Re: How to model searchable properties of an entity

From: Karel Miklav <karel_at_inetis.spppambait.com>
Date: Thu, 19 Aug 2004 12:45:58 +0200
Message-ID: <cg20dd0k6s_at_enews1.newsguy.com>


Bernard Peek wrote:
> In message <cfvhht02nig_at_enews3.newsguy.com>, Karel Miklav
> <karel_at_inetis.spppambait.com> writes

>> Bernard Peek wrote:
>>> If you are going to have the metadata evolving as the database is 
>>> built  then there is no alternative but to have highly trained users 
>>> with an  understanding of information management. There are several 
>>> different  ways of using them but this basic requirement is pretty 
>>> much unavoidable.
>>
>> There is an alternative. Although you may not have a clue what the 
>> metadata will look like you can implement its functional categories in 
>> your software. Then require users to categorize the metadata as they 
>> enter it.

>
> As I read it that requires you to know enough about the unknown metadata
> to be able to define categories for it. That's how the Dewey
> classification system works. You define 10 major categories and fit new
> entries into one of them. The problem is that one of these classes is
> always "miscellaneous" and that particular bin always ends up a mess.

Don't know about libraries but people classified things into discrete categories from ancient times. I can see the problem of the "miscellaneous" category but it's just a design error itself. Like if you're trying to patch a roof with a sieve in case some rain might want to come in. If you have some time to waste I suggest reading structural anthropology, i.e. Claude Levi-Strauss.

> The difficulty with handling unknowns is that they are unknown in

No, they're not, theoretically. People can not perceive uncategorizable miracles (ok, some can :), neither they can enter them into information systems. I'm talking about functional categories and if you don't know what is the use of data you're collecting then why bother, there are random generators.

> unknown ways. You can't be certain that the next one will fit into any
> of the categories that you created. Unless, of course, you make the
> system completely generic. That's the e-a-v system.

Not generic but abstract. Users can not define functional categories on the fly, unless you're one level more abstract then them.

>> Principle is the same, but if lots of folks will use this application 
>> it might be easier to add some code than reeducate the whole industry.

>
> You can take steps to reduce the effort needed but you can't eliminate
> the work completely and all of the users still need to be fully trained.
> If you have the ten Dewey classes you still have to train all of your
> librarians about boundary conditions. For instance which class does
> computing belong in? You have a choice of mathematics, electronics,
> logic or miscellaneous.

That is simple, because users already know their domain, but are usually unwilling to put off things for some loooong lasting IT torture. I'm suggesting to build some abstraction rather than drop users on bare bones of the database. But hey, we al know that :)

Regards,
Karel Miklav Received on Thu Aug 19 2004 - 12:45:58 CEST

Original text of this message