[Top] [All Lists]

Re: MS vs. pop and imap

2004-06-01 10:34:05


One observation...

While I tend to agree with Ned that looking at the history is not particularly useful, I suggest that your summary of the history is not correct either (it certainly isn't consistent with my recollections). In the stage before the idea of a Message Store was more or less explicitly slid into the architecture and we still spoke of an MTA handing off a message to an MUA, POP and IMAP were usually described as "split-UA" arrangements. I.e., with "traditional" access mechanisms, the UA had a user-facing component and a component that pulled the message from wherever (or in whatever state) the MTA had delivered it. With POP and IMAP, that MUA was divided into two parts, one that co-existed on the same host with the MTA's delivery mechanism and served out data to the the other one. The latter sat on a client machine and requested and/or received and/or accepted those data.

Now, even with the introduction of Mail Delivery Agents and Message Stores, that "split-UA" model may still be somewhat useful. I note that we don't make any attempt to distinguish among single-component MTAs (of the emacs rmail command, or the bin/mail variety, or up the complexity scale from it) depending on whether they facilitate leaving messages in the mail store after reading or require either deleting them or moving them somewhere and that, similarly, we don't make distinctions based on whether those MUAs can modify the stored message with annotations, "answered" flags, etc. I think that, if one tries to contend that POP and IMAP are different sorts of beasts, then one needs to start reclassifying the MUAs that don't involve client-server splits, and that would get us quickly into a silly state.

Using similar logic, I think I'd join others in discouraging hair-splitting about what is and is not a legitimate mail store. The reality is that we have fairly few firm boundaries in this business (as someone, I think Ned, pointed out, the concept of "final delivery" is one of them, even if one wants to quibble about whether it should have been called "final") and that the sort of model you are trying to develop is probably best seen as descriptive, rather than normative. If you try to make it normative, I think there are big problems ahead.


<Prev in Thread] Current Thread [Next in Thread>