>Yes, that's true, but let me say again, to support this offset >notion, the FS must enforce it. I would like to think that, for >example, the sum of the part sizes plus header would add up to >the total RFC822.SIZE of the message, but some server implementation >is bound to falsify that assumption, and you have to make up for it. >You will need the embedded MIME headers and separators, and it's up >to you to make sure that the offsets you compute before you've seen >the data leave room for that. An idea : Always fetch multipart messages part-by-part and keep storing the offset after each part is fetched.