Re: Indexes and Logical design

"David Cressey" <david.cressey_at_earthlink.net> wrote in message news:VsHUe.9887$9i4.8700_at_newsread2.news.atl.earthlink.net...
>
> "VC" <boston103_at_hotmail.com> wrote in message
> news:GKqdnRE4MfQ4gb7eRVn-sQ_at_comcast.com...
>>
>> "David Cressey" <david.cressey_at_earthlink.net> wrote in message
>> >If we're going to take away the pointers they love
>> > to play with, we'd better throw them a bone. Indexes fill that bill.
>> >
>>
>> I do not understand the above. Are you talking about some psycological
>> problems that can be settled if the hypothetical, non-database,
> programmer
>> would know tha some table has indexes ?
>
> It has to do with the real world. Nothing that needs to concern you, VC.

Very cute.

>
>> You are right of course that constraints are part of the logical model.
>> However, the unique index is not. Such index is just an implementation
>> vehicle for a unique constraint. Besides, in Oracle (with which you are
>> apparently familiar) , a unique constraint can be enforced via a
> non-unique
>> index. This piece of trivia should make it clear that an index, either
>> unique or non-unique, is just a trick to improve performance. One can
>> easily imagine a unique constraint enforcement without any index
> whatsoever
>> although such enforcement would be impractical.
>
> Not true. In DEC Rdb/VMS a unique constraint can be declared without
> creating an index, if you want to.

For toy tables probably. In 'real world', no.

> There are actually cases, though rare, where that's the right thing to do.

For example ?

>
>>
>> >
>> > Fourth, the DEC Rdb/VMS command "show table" shows the indexes as
> part
>> > of
>> > its display.
>>
>> Surely, you realize that it's just an impementation pecularity, a
>> convenient tool for the DBA.
>
> Life is full of "implementation peculiarities". So far you've dismissed
> Oracle, Rdb, and Data Architect
> peculiarities as unimportant. What's left?
>

What kind of argument in favour of indexes being part of logical model is that ? Are you familiar with the notion of relevancy ?

>> Well, as a database designer, you convert your conceptual
>> simplification
>> of the 'real world' into a logical model and then you think how to
> implement
>> this model using various tools at your disposal (indexes being one of
>> them). Such impementation would be your physical model.
>>
>
> Nope. There is more than one logical model that corresponds to the same
> conceptual model. You make choices based on the probable consequences
> downstream, after implementation. Some of this is hunch work, and not
> strictly mechanical.
>

What kind of argument in favour of indexes being part of logical model is that ? Are you familiar with the notion of relevancy ?

> Again, it's from the real world.
>
>
>
>
>
>
Received on Sun Sep 11 2005 - 00:30:27 CEST