RE: Where is IBM?

> From: Walden H. Leverich III
>
> Now we're getting into a discussion of query optimizer design. I'm
> sure we can agree that _both_ Microsoft and IBM have experts that
> are _MUCH_ more versed in how this works than you or I.

Hah!  You take much for granted, oh Great MicroChief, when you assume that
we can agree!  <grin>  Seriously, you're of course right on this.


> However, I would expect that both systems evaluate the where clause,
> open the index, lookup the keyvalue in the index, get the offset
> into the dataspace and then get the row from that offset. Simple
> enough.

That works fine when you have an index, but not so well when you don't.  And
that's what the address space issue is all about - creating queries.  You
might want to re-read those two links.


> Single-Level store or not, a database is just a byte stream.

Ugh.  Only in your world.  In my world, a database is contiguous blocks of
fixed-length records stored physically, with indices over those record built
logically.  You don't create an EVI over a byte stream.

Now, there are exceptions, including VARCHAR and BLOB/CLOB, but those are
exceptions rather than the rule (although I really like VARCHAR for certain
things!).

Anyway, as you have alluded to, I'm more than likely arguing way over my
head.  The database deities are undoubtedly chortling as they read my
diatribes, so this is probably a reasonable place to admit ignorance and
move on <smile>.

Joe