Handling record locks

["York, Albert" <albert.york@xxxxxxxxxxxxxx>]

As we all know, problems with record locks have been around since day one.
Resolving them isn't easy and someone is going to lose out. 

I have an idea that I would like to present for discussion.  

Perhaps if we added another extension to the READ and CHAIN op codes.
Something like (T), which would be a 'temporary' lock.

If another job already has a T-lock an error would be returned, unless a
certain period of time has elapsed since they read the record (like when
someone leaves a screen up). In that case then the previous job would lose
the 'lock' and you would get the record. 

It you're unable to 'lock' the record you could then present a screen to the
operator giving them an option to force a timeout on the other job (perhaps
this requires a supervisor override).

When you try to update the record if you lost the T-lock to another job an
error message is returned. You then display a 'sorry' message to the
operator. 

If you close the file or release the record, or if your job ends,  then the
T-lock goes away. 

Albert York