Comment by cryptonector
3 years ago
The fundamental problem is that `UID`s in IMAP kinda suck because assigning persistent, unique IDs to emails in a store is a hard problem because doing that for mbox- or maildir-like stores is hard because those predated any notion of remote email access protocols.
Thus in practice IMAP servers generally assign `UID`s ephemerally per-session, which means that clients can't rely on the stability of `UID`s, which means that clients have to re-obtain `UID`s before operating on emails via IMAP even if they have cached those emails locally. `UIDVALIDITY` exists to help clients cache and invalidate `UID`s. The RFC has text about this.
A bridge from IMAP to something else (which is basically what every IMAP server ever is) needs to deal with this. To make `UID`s stable requires keeping state.
Clients should really not assume stable `UID`s. Instead clients should `SEARCH` or list to get [temporarily] valid `UID`s then use those to delete etc.
Why is this complex…just take a sha3 hash of the email body + timestamp, then b58 it or format any way you want. Instant uuid per email.
Yes, that's what I say elsewhere. Digest some of the headers (not all) and the body. If you do it right you can even use that for dedup.