> Alternatively, have a share count in the index file along with an > "expunged" flag. As each process that has the mailbox > selected notes the > message as being expunged, decrement the share count. When the share > count reaches zero, it is then safe to delete the message. > > The update on the index file to clean up the zero-share-count expunged > message can be deferred until some process gets an exclusive > lock on it.
I'm afraid that I can see issues with this and would like to know how you handle it. Client A connects and does a long fetch of all the flags (takes say 20s) Client B connects and does a store 1:* /deleted followed by expunge meanwhile Client C connect and does a store 1:* /seen Now the mailbox is locked for 20s and all 3 command sequences are concurrent (started at the same time) During that time the messages have been flagged as deleted, untagged and expunged Questions? 1. are the messages flagged or not? 2. Are they expunged? 3. what does the fetch return for the expunged messages? Regards Richard Bang Floosietek Ltd [EMAIL PROTECTED] http://www.floosietek.com
