Re: Relational Model and Search Engines?

In message <PIqdnUyLPYBzRgbdRVn-iQ_at_comcast.com>, Laconic2 <laconic2_at_comcast.net> writes
>
>"Anthony W. Youngman" <wol_at_thewolery.demon.co.uk> wrote in message
>news:53FJmwHEI9mAFweF_at_thewolery.demon.co.uk...
>
>> Which is why MultiValue, Tree, and most other models that muddle up
>> theory and implementation will kick the sh*te out of relational when you
>> actually run them in the real world ...
>
>I'm not sure quite what you meant by this. What you run in the real world,
>isn't a model. It's the real thing.

No it isn't. It's an implementation of the model. Which isn't the same thing.

Let's take an accounts department. "The real thing". You analyze their work flow, the *information* they work with, and reduce it all to *data* and a model of the department. You now implement the *model* in a dbms.
>
>I'm not familiar enough with any MV or Hierarchical products to make the
>comparison. But I think it's fair to compare VAX DBMS (A CODASYL database
>product) with VAX Rdb/VMS (a relational database product). They both had
>a common layer of code, called KODA, way below the surface. Going back
>about 20 years, in the heydey of the two products, a VAX DBMS application
>would "seriously outperform" a VAX Rdb/VMS application, given comparable
>loads and resources. I wouldn't go so far as to say "kick the crap out of".
>
>Why? Once you do all the tuning and optimization, one of them gives you a
>pointer, and the other gives you a foreign key, and an index, from which
>you can get one or more pointers. The pointer is faster. Actually,
>Rdb/VMS had a datatype, called DB:KEY which you could use to "cheat". You
>could use a column of this type to implement your own graph when you wanted
>to, provided you were willing to pay the price.

To give you an idea of why MV is so fast, pretty much effectively the foreign key is the pointer ... and we don't normally bother optimising or tuning, either - we don't need to ...
>
>But don't tell the bishops of the first relational church. They don't know
>we're here. (Reference to the old joke about heaven).
>
:-) What I'm getting at, is that if you hit a *thorny* problem *implementing* the model, the theorists will say "not our problem". And if (sod's law saying "of course you will") you hit a knapsack-type problem, the theorists will walk away.

Cryptographers love knapsack problems - brute force takes forever and trying to solve them intelligently takes almost as long :-)

Cheers,
Wol

-- 
Anthony W. Youngman - wol at thewolery dot demon dot co dot uk
HEX wondered how much he should tell the Wizards. He felt it would not be a
good idea to burden them with too much input. Hex always thought of his reports
as Lies-to-People.
The Science of Discworld : (c) Terry Pratchett 1999