Re: Object-relational impedence

Ed Prochak wrote:

> On Mar 19, 1:15 pm, S Perryman <q..._at_q.com> wrote:

>>A standards body basically defined resource types of the form :

>>type T
>>{
>> Attr1
>> Attr2
>> // etc
>>
>>}

>>Attr1 = string
>>Attr2 = Sequence<SomeNastyRecordType>

>>There were 100+ resource types, and at least 100+ attr types.

>>Which types had which attrs, and whether they were mandatory (belonging
>>to all instances of a types) , or optional (present on a per-instance
>>basis) was defined by the standards body (for the perversely-interested,

>>This is what the developers were faced with.
>>Store the info for each resource instance (and meta-info to allow the
>>generation/processing of outgoing/incoming messages requesting/retrieving/
>>setting the attr values) .

>>Their mandated impl technology : SQL RDBMS.

> So they designed the data model to match the Standards document? IOW
> they did not do any design, but just tried to implement the standard
> in the RDBMS directly, right?

Wrong.

> My first impression (coming in late to this thread) is that is what
> happened. And that is clearly an EAV design. Given a standard like
> that I think I would quickly decide to split types into different
> entities.

As did they (except for attrs that were multi-valued of course) .

 > [ conclusions drawn from speculation and not fact - snipped ... ]

Regards,
Steven Perryman Received on Thu Mar 20 2008 - 10:19:38 CET