Re: [GENERAL] Joins and links

Bruce,

It is amazing when you get responses written this fast (so that the
reponse arrives before the copy of the message from the list).

> In fact, you were mentioning that inserting into the middle is slow, but
> that sequential adding to the end is good,

Yes this is what I was told about the way MS SQL Server does clustering.

> but in fact, heap already does this, doesn't it?

heap? I am not sure what you mean.

> I guess if you only add occasionally, it is OK.
> Also, our no-over-write table structure had a tendency to mess up that
> ordering because updated rows do not go into the same place as the
> original row.

I have just been thinking a bit more and have realised that the
multi-generational architecture of 6.5 (which I have used in Interbase)
means that probably both clustering (in thr dynamic sense) and full
record number support as request by Leon are impractical.

It seems to me that record number relationships will fail completely if
there can be more than one version of a record. (well even if they are
forced to work they will lose some/all of their speed advantage).

Dynamically clustered indexes might still work but unless tables are
appended to only with no inserts or updates then maintainig the table in
index order when there can be multiple version of each row would be very
slow.

Dave

--
David Warnock
Sundayta Ltd