Re: Semistructured data

Sampo Syreeni wrote:

> On 2006-10-26, Marshall wrote:
>

>>> [...] the final, disjunction aware database ends up being emulated on 
>>> top of the RM.
>>
>> Can you give an example?

[long-winded straw man snipped]

> The problem is that the database has no idea about what this all means.

Sets of facts don't have ideas and don't ascribe meanings. Do you have a point?

> When asked, it can easily list all of the statements and formulas, but
> in order to get out only the true facts, you'll need extensive reasoning
> capability on top.

Um, you mean like a query?

  The data model that lives inside the database is far
> more complex than the fragment of propositional logic the RM captures,
> so the RM is mostly blind to it and is only used to emulate the real thing.

That is a remarkable assertion--all the more remarkable for its complete lack of support or evidence.

> Finally, the database isn't really even equipped to deal with the
> pattern of queries that result.

Databases aren't equipped to deal with queries--that's what data management systems do.

[handwaving about combinatorial explosions snipped]

Then there are going to be predicates describing incompatible
> medications, side effects, afterillnesses, combined risk factors and
> whathaveyou. Those translate into path queries and mutually recursive
> transitive joins,

Path query? Transitive join? What utter nonsense. If recursion and transitive closure facilitate expressing the queries, then express the queries using recursion and transitive closure.

  with intermediate results that are going to be huge

Bullshit.

> unless the database actually understands the logic and can prune the
> search intelligently.

Sets of facts don't understand things. However, the manipulation function of the dbms uses symbolic manipulation of formal logic to avoid unecessary work.

[remaining nonsense snipped]

Sampo, you have no intellectual honesty. Your posts are long-winded and almost entirely devoid of sensible content. Plonk. Received on Fri Oct 27 2006 - 15:32:07 CEST