Jörg,

> When i read this a couple of questions dive up.
>
> 1. How large is large?
> 100?
> 1000?
> 10000?
> 60000?

With a 128K line, you can transfer a maximum of 16K per second
(uncompressed).

If you have larges rows with (e. g.) 4K and you fetch 30 rows (= 120 K),
fetching will take at least 8 seconds.
You can easily use this example for your rows/tables.

Generally, 'large' is if it takes too long to fetch the results.

> 2. At which size should I use a blob instead of a VARCHAR?

There is no general hint for this, but your application must be optimized
for slow transfers, so analyze your tables and calculate the time a row
needs
to be fetched. The fetch time of rows with large VARCHARS can be reduced
when
you use blobs because blobs are only fetched when they are effectively
needed -
this saves a lot of time.
Please also stick to Geoff's comments.
There is a good comparison of blobs ands varchars at
http://www.volny.cz/iprenosil/interbase/ .
Click at "to blob or not to blob".

> 3. Why is a blob field better than a large VARCHAR?

Blobs are only fetched when they are _really_ needed.



HTH,

Christian