Leyne, Sean wrote:

>>...isn't the better solution a SQL analog of the isc_unwind_request
>>
>>
>call?
>
>Yes.
>
>
>
>

OK, let's look at that. There are two problems with
isc_unwind_request. One is that it doesn't work on SQL statements, the
other is that it doesn't actually work.

New model:

1. isc_statement_info makes a available a request "token"
2. The "request token" is used in a separate connection to kill a
running request. Dpb parameter? New statement? Don't care --
anything would work.
3. A polite request to kill an active request with a given token is
honored as long as the attachment accounts were the same.

This doesn't require a change to the API or plumbing and could be
implemented in an afternoon.


[Non-text portions of this message have been removed]