Thank You very much, everyone for helping me out. Much appreciated
that i've got so much feedback, even tho' i'm far from finished
myself with researching the problem.

I was hoping that i had run into a well known problem with some wrong
setting in one of the components or some interbase setup that caused
the problems. As the program is mostly programmed on the server and
in the forms (yuck!) I'll probably fix it in the end by changing some
properties here and there. I'll never get the hang of that kind of
programming ;-)

Some questions have risen in the group, and I have been at the
customer, and I've found out more :

- they claim now, that they can make a workstation fast in about 1 hr
(it was a day before) if they use the grid intensively enough
(searching a bit, page up a bit page down a bit, arrow here arrow
there). That should make it possible for me to duplicate the problem
programatically here, which i will try to do right away.

- when the grid is slow, all it takes is to close the application on
the workstation, then open it again - then everything is fast.

- there can be one slow workstation and several fast ones at the same
time on the same server (i haven't verified this 100% but they were
sure that was the case). There can also be several slow workstations
at one time.

- not sure if all grids in the application slow down, or only the one
that has been used a lot. I strongly suspect it is only the one they
use a lot.


When i have set up a system here, that can fail on demand, i'll take
a look at the sql traffic between client and server. i'll dump a
minute when things are fast, and a minute when things are slow. That
should give me some insight. especially if the two dumps are similar
to each other. (then i has to be transactions or memory leak or
something)

I will also make something that can time a set of pseudy random
operations on the grid, to measure the results of different ways of
attacking the problem.

You have all been of great help and inspiration, and i'm very
impressed that anyone cared reply, as my initial bug report was
indeed without much detail.

I'll be back with the results when the problem is fixed, or when i
give up trying (memory, transactions, autofetchall)

Btw. whoever guessed it's an old paradox application that was
converted is right on mark!. it is!. i'm sure ibobjects is the only
reason it works at all right now. Must admit that the dataaware
controls and the ibobject "framework" works far better than i had
ever imagined possible. I also agree that the basic user interface
structure is not very C/S friendly in any way. However, i plan not to
change that part of the application unless the customer throws some
cash at the project. A derived question :

if one was to loose the whereless selects, one should be able to
present the user with a screenfull of records, and easy access to the
prior screenfull and the next screenfull (like search results on the
web)

How do you do something similar to a search for all somethings, with
a Next page, and a Prev page, like on a webpage. You don't know the
boundaries for the search for the next batch, only that they (if it
was names) are the next 20 records, after the 3rd "Jason". (if sort
on lastname and last 3 lastnames on screen was "Jason").



Dennis Fantoni