From: Robert Ramey (ramey_at_[hidden])
Date: 2004-01-03 11:58:48
David Abrahams wrote
>Also, as long as you're going to save the unmodified data before
>beginning an operation, you might as well use the undo data for
>exception recovery. In that case, you probably want a system for
>restoring the data which can't throw an exception. Deserialization is
>unlikely to fit the bill.
The serialization package under discussion only throws an
exception if the archive has been corrupted some way. So
any exception during de-serialization would either be a
non-recoverable programmer error or an exception invoked
by some other system such as I/o which might be recoverable.
I don't see why the deserialization process can't throw an
an exception an if that exception is something recoverable
appropriate action can't be taken.
That is, I don't see that de-serialization raises any issues
regarding exceptions that are not raised by exceptions
Boost list run by bdawes at acm.org, gregod at cs.rpi.edu, cpdaniel at pacbell.net, john at johnmaddock.co.uk