Hi David

Even in 6.4 I'm pretty sure that the insert into heap should fly
and give consistent performance no matter how full it is.

> Was a long time ago now, but I don't recall any indexes being
> involved.

May have been associated with a constraint.

What about an insert rule on the heap table? Something that
says An Insert into A means all this extra work on B and C and D ...Any
of which could have gone into overflow. Or have required an aggregate
calculation...
>
> What about:
> - constant file extending?;

I would have thought that would be a constant. And not to lead to
progressivly worse performance. The default extent of 16k is going to
have much the same performance on each attempt. I suppose if the
disks were extremely full and fragmented that this might be a problem.

> - an ingres bug (last resort of the desperate).

Always possible.
>
> Have to say, I'm happy to find I've been wrong!
>
Who said I'm right? According to the missus I havent been right
for years. If a man walks into the forest and says something and there
is no one around to hear. Is he still just as wrong?

Marty
--
Random Titus Quote #23:
Revenge is good.
It's what seperates us from animals and hippies.