> Alan McDonald wrote:
> >> We just use GUIDs here for IDs (as our software runs in a 3-tier
> >> environment
> >> its better, because we can generate the ID on the client
> side), and aside
> >> from the little problems mentioned by Alan (which aren't real
> problems) we
> >> don't have any complaints.
> >>
> >
> > you're aware that GUIDs provide poor indexing with FB's index
> trees don't
> > you?
> > Are you happy with your PK index selectivity?
> > Alan
> >
>
> Well.. once it's a PK it could be only unique and so forth the best
> selectivity available :-)
>
> I think you have written wrong and meant "size" or "compactness" instead
> of selectivity.
>
> The number of index entries on a page will be of course small when one
> uses a UUID/GUID instead of a integer. I don't have numbers about how
> compression goes on a series of UUID/GUID values (since it's "random"
> the redundant data should be small and has poor compression). Maybe the
> compression ratio is terrible, maybe not so bad as one expects, I don't
> know... :-)
>
> I think the OP should give a try and populate a table with UUID/GUID
> values and use gstat to see the average size of the index entries
> compared to integer fields.
>
> see you !
>
> --
> Alexandre Benson Smith