At 01:21 PM 16/10/2005 +0000, you wrote:

>I read the detail again and perhaps I have a slightly better
>understanding but I'm stil not sure.

You seem to. :-)

>Suppose I set AllowCheckAOT to
>60 and Attempt to 60.

If Attempt kicks in first, then after 60 seconds' idle time (= user
painting her nails?) Attempt will begin the FetchAll. AllowCheckOAT won't
be counting if there's a FetchAll going on, since the app is not idle in
that case. So you can assume (I think) that if AllowCheckOAT didn't get
there first, it will be 60 seconds after Attempt has fetched the last
record that AllowCheckOAT will return a hit and start to wrap everything
up, i.e. with these settings there would need to be at least two minutes of
idle time before the bird is likely to flee the coop.

>Does this mean that IBO won't check for active
>transactions until 60 seconds have elapsed and then will attempt to
>fetch caFetchAll datasets to EOF for another 60 seconds before
>closing the transaction?

Why not set it up and test what happens?

>Is the time set in AllowCheckAOT start when
>the transaction is started or from the last activity for that
>transaction?

The last activity in that transaction, which might be the start of the
transaction (if no DML was posted) otherwise the last call to Post or even
just a keystroke or mouse action that the application can
detect. Remember, it's checking for activity that would hold back the
advance of the OAT, so that's going to be *any* started read/write
transaction that has open cursors or pending DML. In the case of
Autocommit and/or CommitRetaining transactions that have open datasets or
prepared statements, that means all transactions, of course.

Helen