Martin,

> Isn't is easier/more logical to release the lock as soon as the last
> transaction that is using the file ends?

I don't know about 'easier' -- but I have my doubts.

As for 'more logical', that depends on your view/expectation of
persistence. Consider:

- User 1 defines external table.

- User 1 accesses external table and committs his work. And proceeds to
perform other tasks.

- User 2 (as Mark O described) -- a wally of a SYSDBA -- deletes the
source table from the server.

- User 1 start another transaction and tries to re-access the external
table -- receives an error message. Wonders that the @#^&&#$ happened?


Personally, I would expect that if I'm able to access an external table
once, it should be available for the remainder of my session.

I think that the whole issue about locking external files (actually
keeping the handle open) is a red herring. The real issue that we need
better tool to provide fast(!) server based data import (and export)
functions.


Sean