RE: A suggestion for RFC1939
Laurence Lundblade <[email protected]>
| Newsgroups | gmane.ietf.pop3ext |
|---|---|
| Message-ID | <[email protected]> |
At 09:32 PM 3/27/00 +0530, Faheem Ahmed Khan wrote: >I agree that the UIDL to the message number mapping doesn't change during >one session, but it can change between consecutive sessions. > >For example: >there are 100 mails in the mailbox. >I have retreived 95 and my connection to the mail server is terminated >abruptly. Though I have issued DELE after every RETR my messages still lie >in the mail box, and in my next connection to the mail server I have to >match the UIDL of the message already read by me with every message in the >mailbox before retreiving new ones. Yes, you do have to do this as things stand now. >BUT, if I could use UIDL with the DELE >command I wouldn't have had bothered about that and issued a batch delete >for all the messages already retreived and still be sure that I have not >deleted an unread message. Wouldn't that save me from comparing the >MessageID of every message on the server with the ones already read? It would save you comparing. BUT, your mail client would only be able to talk to new servers with this new feature. It will take five years or more before all the servers are upgraded to allows UIDL with DELE. This is not an exaggeration! People just don't upgrade software that fast. In the mean time your client software would be sort of useless. There's actually many things like this in Internet protocols that could be redone much better now with hindsight (RFC-822 address syntax for example, or the chattiness of SMTP), but backwards compatibility stops us from changing things. The deployed base is huge. In your case of POP and UIDL the inconvenience to programmers like you and I is relatively small. All you have to do is build a UIDL to message number mapping table when you start your POP session. In other cases the messes are much bigger. Basically I don't think you're going to get very far pressing this. LL >-Faheem. > >-----Original Message----- >From: [email protected] [mailto:[email protected]] >Sent: 27 March 2000 09:09 >To: Faheem Ahmed Khan >Subject: Re: A suggestion for RFC1939 > > > > > > Faheem Ahmed Khan wrote: > > I don't understand why RFC1939 restricts to using message number with > > the DELE command. Using the messageID (UIDL) with the DELE command > > would be a lot more useful where the user agent can be sure of which > > message it is deleting. > >I don't see how it could be more useful, since the client needs to know >the message numbers for the other commands, such as RETR. The UIDL to >message number mapping never changes during a session, so I don't see >how the client could be "more sure." > >At the time UIDL was created, we were trying hard to limit the number >and scope of changes to the protocol. The fewer changes, the fewer >things for implementations to get wrong.