Re: foundations of relational theory? - some references for the truly starving

>So is Microsoft Access and Fox Pro and Paradox. Who gives a damn about
>these ?
But you can't run concurrent systems supporting hundreds (thousands?) of users on Access, Fox Pro or Paradox.
You can on Pick (current variants anyway).

>Without a scientific work describing what the features are, what are the
> integrity consttraints, expressive power, is it relationally complete,
>is the language declarative, etc, etc, I'm sorry but I can only
>disregard your model as a matter of principle, just the same that we
>have to disregard anybody who claims that he has a perpetuum mobile, or
>he has a valid pyramidal scheme.

>Sure. If PickDB offers me the same relational features as say, Oracle
>I'm ready to consider it for the next project. If all it offers is
>instead a development toolkit a la FoxPro, or no integrity constraints
>and limited declarative query , or problematic concurrency control, or
>doesn't have convenient or production ready language bindings for Java,
>C#, C++, Visual Basic, then forget about it, right away.

Pick variants from IBM (UniVerse and UniData) offer all these things.

>How can I judge these for myself ? I am not going to read it from
>marketing brochure, or am I ?
Are you confusing scientific documents about the relational model with descriptions of actual implementations.
Can you provide me with a scientific document describing Oracle or DB2 specifically?

>I am not going to download and setup ellaborate load tests, and learn the
stupid damn thing, or am I ?
>This attitude of "try and you'll be enlightened" is completely idiotic
But quid pro quo, why should we try Oracle or DB2 or SQL Server?

>If it was that smart Pick Db can blow out at least TPC-W and TPC-H which
>are sufficiently technology neutral, at price performance comparison.
Having read the conditions for the TPC benchmarks, the particular benefit of pick (nested relations) is specifically excluded from use (since the table structure must be implemente exactly as described). Should we do this, although the benchamarks would work I doubt that performance would be spectacular (probably not horrible though).

If we could nest some of the structure (and alter the SQL queries appropriately to UNNEST) then the TPC benchmarks would be worthwile assesments of Universe or Unidata.

Craig Received on Wed Oct 22 2003 - 05:51:59 CEST