On Mon, 5 Jan 2004, Timo Sirainen wrote: > > At least the task is a little bit easier if you store UIDs with the > > message, rather than having to infer "same message" from message contents. > Several POP3 servers get UID by getting MD5 sum of some part of message, so I > think that's also quite safe to use.
That is a form of inferring "same message" from message contents. You have to be careful to distinguish between static message contents and dynamic contents (such as Status: lines). Also, you can't use MD5-derived UIDs as an IMAP UID. IMAP UIDs are 32-bit and are strictly ascending. So, although MD5-derived UIDs can be used there are some pitfalls, you still have to have separate IMAP UIDs, and the IMAP UIDs are sufficient for the purpose without having to use MD5-derived UIDs. > I think most modern MUAs also behave pretty > nicely. IMHO, MUAs only behave nicely with each other if they use the same code to access the file data (such as the c-client library). Otherwise, you have to assume worst-case behavior to be safe. > mbx works only with local filesystems and it has very little > application support outside c-client. That's by design; the idea was to discourage other implementations of mbx so that all software that supports mbx uses c-client. That way, these programs all play nice with each other. I actually have had someone tell me that "c-client handles mbx wrong" because of some interoperability problem with a different implementation, and had to explain to that person that c-client's handling of mbx is right by defintion... :-) Not everybody grasps that notion right away. -- Mark -- http://staff.washington.edu/mrc Science does not emerge from voting, party politics, or public debate. Si vis pacem, para bellum.
