Hi Marcus,

nice list you have there...
One of the things that can be measured is the time difference
just using networked tables.

As long as only one machine uses the networked table,
the speed is fine, since the table is already open under
another alias.

If the dbf() is not used, the "use" will take slightly more time.

As soon as another machine (another process on the same machine DOESN'T slow the speed) has the table already open, the first "use" to open the table will take between 10 and 20 times as long. Use again doesn't take the same hit, but is slower as well.

My Assumption: the OS "caches" if running alone in one or more processes and this won't work if the table is already opened in another WS.

That alone can bring widely differing timings to a mix of
"use" and sql-select's.

You mentioned increased network traffic: if the other app
writes data and changes fields which are indexed or deletes
in tables with an index in deleted(), the indes will be flagged
as dirty AFAIK and each workstation has to read the whole index
before any index-based or rushmored'd operation -
which also could introduce a lot of different timings.

And then there are different combinations of server and client OS.

If it's not too urgent, let's have a beer or two at the devcon...

cu

thomas