Date: Sat, 9 Jul 1994 16:47:29 -0400
From: "Steven D. Majewski" <sdm7g@elvis.med.virginia.edu>
Message-Id: <199407092047.AA16745@elvis.med.Virginia.EDU>
To: Mark Crispin <MRC@cac.washington.edu>
Subject: Re: sequence numbers ( was: (long) sketch of proposed imap: URL syntax ... )
On Jul 9, 12:11, Mark Crispin wrote:
>
> I object vehemently to the notion of deprecating sequence numbers.
>
I was by no means suggesting that sequence numbers be deprecated in the
IMAP draft!!
That word was used in reference to their proposed use ( by me,
in the first pass ) in an imap *URL* specification. "deprecate"
was the wrong word - since there is no standard usage we are
deprecating FROM yet. I should have used the simpler phrase:
"to drop sequence numbers from the proposed URL specification"
I've only gotten negative comments on the use of sequence numbers
in a URL, and I largely agree with those comments.
> Sequence numbers serve a useful purpose. They will always be faster than
> UIDs, and are perfectly satisfactory as long as you only use them within a
> single online session. Sequence numbers are the only reasonable way to
> specify concepts such as ``the sixth message'', ``the last 50 messages'',
> ``the first 20 messages'', etc.
You are, of course, correct here - so using the work "mistake" was
clearly a mistake on my part. Sorry. Let me re-phrase that: the
"mistake" was NOT the *inclusion* of sequence numbers, but was the
*omission* of UID's or some similar notion from the original IMAP2 spec.
( An error that is fixed in IMAP4 )
My problem is similar to the "disconnected use" situation. I'm trying
to map between stateful and stateless protocols.
>
> I never advocated presenting sequence numbers to the user. None of the GUI
> clients I have written do so (in spite of considerable pressure).
>
Then I assume you feel that NOT presenting them to the user ( in an IMAP URL )
would be a good continuation of that tradition.
I may have also been guilty, in parts of this thread, of blurring the
distinction between what the server makes visible to the client, and
what the client makes visible to the user. Since I am building a
gateway - this distinction DOES in fact become blurred for me - I don't
directly control the user interface. ( But no excuse for propagating my
confusion! )
But while we're making that distinction clear, I'll reiterate what I
said of the parallel problems of UID's. It would appear that making
them visible to the USER ( as opposed to the client ) would be a
possible source of problems.
They seem to be guaranteed unique as a function of ( server, mailbox ),
persisting across sessions, but NOT across access thru different
servers. Message ID's are also not unique. Even if they are unique
to a message, they may not uniquely identify a single copy of that
message in a particular mailbox. But I suspect that I'm the only one
who finds it a problem to not have a reliable unique indicator of a
particular message in a particular mailbox. Sequence number, UID and
Message-Id all fail in different ways. Message-Id, at least, has the
plus of being less likely to fail "silently" by returning the wrong
message.
- Steve Majewski (804-982-0831) <sdm7g@Virginia.EDU>
- UVA Department of Molecular Physiology and Biological Physics