Re: IMAP URL scheme <draft-newman-url-imap-05.txt>

On Thu, 30 Jan 1997, Larry Masinter wrote:
> 2.1: syntactic compatibility
> The only other URL schemes that access "message in repository" are
> "news" and "nntp". I suppose IMAP doesn't have any way to reference a
> message by its message ID, although NNTP does by its message number
> within a news group.

IMAP UIDs are semanticly similar to news article numbers.  There is no way
to fetch by message ID, although searching by message ID within a folder
is possible (and, of course, more than one message could be returned).  I
could not use the "/" separator which NNTP uses since "/" is a valid
character in a mailbox name and quoting the "/"'s in a mailbox name was
not only unpopular when an earlier proposal was reviewed, but would break
relative URL resolution -- something nntp URLs don't support.

> The section 8 exception for "relative IMAP URLs" is just unacceptable.

Given that (a) my exception matches Tim Berners-Lee belief of how
paramaters should be handled in relative URLs, and (b) the resolved URL
would make no sense without it.  I believe it's reasonable and necessary.

> I'm skeptical that "IMAP4 permits the use of any hierarchy delimiter
> in mailbox names" is a good idea. We avoided this in "file:" and think
> the same arguments apply here.

The IMAP4 protocol does permit any hierarchy delimiter (although only one
per name).  Part of the justification for this was so that newsgroups
could be represented with their canonical delimiter ".".  I note that the
news and nntp URLs don't require that "." be mapped to "/" and my URL
proposal follows that example.

> 2.2: is the definition really complete? I guess I don't know enough
> about IMAP to know whether you can read this spec and write the
> program that translates IMAP urls to IMAP calls. Doing this for most
> protocols is really hard.

I believe so.  My test was the examples section.  I can't know for sure
until someone other than me does it.  Everyone I've talked to is waiting
for this to be published as an RFC before starting :-(

> "2.2.4 Definition of URL schemes not associated with data resources"
> 
> Do IMAP URLs reference data objects always? Even the search mailboxes?
> What's the definition of the interaction? 

It's very similar in concept to the ftp URL, actually.  Some forms refer
to distinct objects, others (including the search) refer to an explicit
list of objects.  Ftp URL doesn't go into details about what information
to include in the list of objects, and I followed that lead.  I consider
this a feature, since different IMAP readers show different sets of flags
and headers in the summary list of messages.

> "2.3.1 Proxy into HTTP/HTML"
> 
> Do you have an HTTP<->IMAP gateway?

No.  Is this a requirement for proposed standard?  If so, I think it's
an excessive requirement.  It would be a fine requirement for draft
standard status.  I know of know reason why such a gateway couldn't be
built, with the exception that HTTP's authentication framework is inferior
to that in IMAP.

> Now that draft-bradner-key-words is on the way to RFCdom, I'd like to
> see new RFCs reference it for the definitions of MUST, SHOULD, rather
> than repeat them.

I have no objections changing it to a reference once there is an assigned
RFC number.  This could even be done in cooperation with the RFC editor
since it's an editorial change only.