> Let me ask a question about Keylinks, if I got it right, the KeyLinks

should

> refer to a field which is not necessarily a primary key of a table but
> rather a field which uniquely identifies a record in the table.

Doesn't have to be a primary key but it does have to be unique for the data
being queried from. That's an important distinction between data queried
from and data returned into the buffers. The KeyLinks columns are used to
uniquely identify data on the server as well as data in the buffer.

> What about tables that don't have such field (I have tables which need 2-3
> primary keys to uniquely identify a record...)?

You can have more than one column in the KeyLinks. Though, it will consume
more memory in the buffers because they are not trimmed and compressed. I
suggest you use a surrogate key if possible.

> Letting the KeyLinksAutoDefine to True should work fine, shouldn't it?

Depends.

PS. Please ask questions in an appropriately labelled message. One per topic
unless somehow inter-related.

Jason