Well, depending on the OS, it might clear itself up after a short time, like ENFILE. So the server *might* be able to answer the next request.
Yes, assuming you trust both your code and any libraries used to recover from ENFILE. Pardon my lack of faith, but I don't trust those libraries to be tested against low-frequency errors like ENFILE.
OTOH, in the case of such resource-shortage errors, dropping the connection and letting the client reconnect would make it more likely that the problem would be cleared up by then. So I guess I'll do that for those cases.
But if a particular message is inaccessible (due to wrong permissions, or being deleted (possibly by a non-IMAP agent), etc.), then NO is appropriate, since the server can still answer other requests, right?
All that sounds sensible to me.
Arnt
