Re: relational reasoning -- why two tables and not one?

From: Bob Badour <bbadour_at_pei.sympatico.ca>
Date: Thu, 15 Oct 2009 23:41:22 -0300
Message-ID: <4ad7dd58$0$23757$9a566e8b_at_news.aliant.net>


paul c wrote:

> Clifford Heath wrote:
> 

>> paul c wrote:
>> ...
>> I hope the SAP example shows that's simply not the case. 500 tables is
>> considered to be medium-sized in my experience.
>>
>> Last year I used CQL to model motor vehicle insurance claims. The model
>> had 100 nouns, though the database was only 18 tables. It did not include
>> most of the complexity of motor vehicle insurance, did not model the
>> policy,
>> underwriting, insurance history, nor many other facets; and this
>> organisation
>> handled more than twenty other types of insurance.
> 
> Thanks for that prompt, SAP was one of the products I had in mind.  I'm 
> very curious to what extent SAP uses views.  I gather that it runs on 
> SQL server, Oracle and perhaps other dbms'.  Does it use some subterfuge 
> to update/insert/delete to/from views?  Does it implement its own 
> integrity mechanisms to get around the various inadequacies of those dbms'?

SAP foregoes any and all dbms constraint checking. It implements its own flavor of SQL that gets translated to the underlying dbms. Views? Bwa ha ha ha ha... let me regain my composure. No, they don't use views, and anything that even remotely smells like database access goes through some function call or another. Updates generally get queued in some sort of unindexed or lightly indexed table where inserts are fast and then processed after the fact. SAP very heavily overloads tables and the side-effects can be staggering. Received on Fri Oct 16 2009 - 04:41:22 CEST

Original text of this message