Date: Wed, 6 Jul 1994 10:03:43 -0400
From: "Steven D. Majewski" <sdm7g@elvis.med.virginia.edu>
Message-Id: <199407061403.AA14923@elvis.med.Virginia.EDU>
To: Wallace Colyer <wally+@cmu.edu>, imap@cac.washington.edu
Subject: Re: (long) sketch of proposed imap: URL syntax and semantics
On Jul 6, 8:44, Wallace Colyer wrote:
>
> > Sequence number and Message-id: *are* usually available and viewable.
>
>
> I use Pine regularly and have my options set so that the mailbox is
> reverse sorted at startup. The sequence numbers displayed start with
> the most recent message in the mailbox. If I were to blindly follow
> instructions for making a URL from these sequence numbers the wrong
> message would come up consistantly.
>
> Clients do not consistantly show IMAP sequence numbers to users.
>
Good point.
However, sequence numbers ARE available (even if some care must
be taken to get then right), and UID's are not, and some servers
do not yet support access by UID.
So I would still maintain that:
The preferred selector for internally/auto-magically generated imap
URL's should be UID (if possible).
The preferred selector for humanly/interactively composed imap URL's
should be the mid: URL scheme when defined, and until then, the
"Message_id:" header field can be used as an approximation.
In a pinch ( i.e., no server suppost for UID's ), in some
circumstances, with reasonable care and warnings, sequence
numbers can server as a message selector. ( And all of these
objections should be collected in a paragraph explaining why
they are legal, but you ought to try to avoid them, if possible.)
Given certain constraints, ( ReadOnly or AppendOnly mailboxes, etc. )
they may in fact be the preferred INTERNAL selector in some cases.
It would be nice, though, if they could be KEPT internal. Is there
a way to "lie" to a WWW client: to use one URL for internal links,
but cause another URL to be displayed on the client ?
URL's are not defined/required to be permanent references ( and
many, from my experience, are not! )
URL's can be improperly constructed, for example: if one forgets
that the http: namespace root is different than the filesystem
root namespace when constructing a reference.
Your example, and John Gradiner Myers "3rd file in the directory"
analogy show that IMAP URL's constructed this way may be a bit more
"fragile" than other protocol examples. But the alternative ( right
now, with existing IMAP2 servers ) seems to be to define imap URL's
so they cannot function as LOCATORS at all ( except for search/query
string selectors ) because UID's are not available on all servers.
[ Well - the *other* alternative is to lie and say that all selectors
*are* UID's , but for servers that don't support UID's, we will map
them to sequence numbers. But I think this is more dangerous than
simply allowing sequence number selectors! ]
-- Steve Majewski (804-982-0831) <sdm7g@Virginia.EDU> --
-- UVA Department of Molecular Physiology and Biological Physics --
-- Box 449 Health Science Center Charlottesville,VA 22908 --
[ "Cognitive Science is where Philosophy goes when it dies ...
if it hasn't been good!" - Jerry Fodor ]