Re: Mulberry gone, now what?

On Fri, 7 Oct 2005, Dave Cridland wrote: 
Polymer should work with any IMAP4rev1 server. It might work with a
plain IMAP4 server, but I'd treat any errors there as low priority.

I don't think that it makes much sense to work with IMAP4 (RFC 1730). IMAP4 was a false start. If you must fall back from IMAP4rev1 (RFC 2060/3501), it would be better to fall back to IMAP2 (RFC 1064/1176).

See http://dave.cridland.net/acap/polymer.html for example where it
says:-

Cool. Mark, see that? My webpage is a more important reference than the
specification. ;-)


:-)

Latitude is definitely not intended to mean the specification is in any
way problematic. IMAP is very highly specified. It does, however,
provide various protocol features which allow server to behave
reasonably differently within some areas, in order to support very
different server-side mail spools. That's not a fault in the protocol,
that's a benefit. (One that has, in many respects, outlived itself).


Agreed.

In
addition, there's exciting choices like multiple commands in progress,
unsolicited reponses at any time, etc.

I never considered these to be particularly important, but there was a small but very loud contigent (now long gone) that screamed for multiple commands in progress.

The original IMAP had commands (with no tags) and four responses: OK, NO, BAD, and *; * carried data, and OK/NO/BAD indicated command completion. RTT avoidance would be done via command pipelining, as in several other protocols.

Supposedly, the need for multiple commands in progress (and thus, for command tagging) was that a SEARCH might take a long while because the operator might have to find the right backup 9-track and mount it; so it was desirable to allow other commands to be executed out of order.

Needless to say, that was never a scenario that existed in real life. I have never seen a server execute commands out of order, much less a client expect command execution out of order.

As for unsolicited responses at any time, besides the BYE response (which is important), there was a hope to avoid the need both for polling and an IDLE command. As it turned out, NAT boxes effectively killed IDLE in its infancy. Long live polling.

These are minor little variations in the gran scheme of things, though.
The vast bulk of IMAP is very tightly specified - in some cases so
tightly specified that virtually nobody implements it correctly
(witness the \Recent flag).


UW and Cyrus do... :-)

My exeperience using mutt with different IMAP servers is that it never
works quite the same with different servers.

Mutt is pretty basic in its IMAP support, so yes, server behaviour -
such as hierarchy - is directly visible to the user.

Yes, and (as I said earlier) that was once considered to be a feature, not a bug.

Like, for instance, servers sending FETCH or EXISTS responses
when no command is in progress, even if they don't support IDLE. It's
legal, but nobody does it. Hardly any clients can cope, either, but
Polymer can, just as it can cope with commands executing out of order.

This is a good example of Postel's robustness principle. The c-client library (thus programs based upon it such as Pine) can cope, but I wouldn't think of actually sending such. The most that UW imapd does in this regard is send unsolicited EXISTS, RECENT, and a FETCH carrying flags, but only while a command is in progress.

Note that the last entry is "Other Servers" - I do know that Polymer
works perfectly well with UW-IMAP, for instance, because UW-IMAP
conforms to RFC3501 etc. I've not put UW-IMAP on that list, because it
happens that I don't have an account on such a server.

Can't you install it on your own system? It supports almost every UNIX variant (certainly all the modern ones). With some effort, you can also run it on Windows.
ftp://ftp.cac.washington.edu/mail/imap.tar.Z
for the current release version, and
ftp://ftp.cac.washington.edu/mail/imap-2005.DEV.tar.Z
for the alpha version of the next release (which adds UIDPLUS support).

-- Mark --

http://staff.washington.edu/mrc
Science does not emerge from voting, party politics, or public debate.
Si vis pacem, para bellum.
.

Relevant Pages

  • RE: RPc server is unavailable since SP1
    ... After these commands run successfully, ... RPc server is unavailable since SP1 ... >> when the member server update certificate you get the error message RPC ... >> interface security settings before the installation of SP1 will be lost. ...
    (microsoft.public.windows.server.sbs)
  • [NT] NetWin DMail Authentication Bypass (dlist.exe) and Format String (dsmtp.exe)
    ... either be used as a small personal mail server or as a 10 Million user ISP ... password hash) when sending the administrative commands. ... the DList server using a numeric hash of the administrative password. ...
    (Securiteam)
  • [NEWS] Netscape Publishing wp-force-auth Command
    ... Remote attackers can easily use the wp-force-auth command to perform brute ... connection with a directory server, which has valid users and passwords. ... Netscape Enterprise has a selection of ?wp-* (Web publishing) commands ...
    (Securiteam)
  • RE: copy permissions from one user to another?
    ... THIS STORED PROCEDURE GENERATES COMMANDS ... -- ADD USER TO SERVER ... -- CREATE TABLE TO HOLD LIST OF USERS IN CURRENT DATABASE ... -- SET COMMAND TO FIND USER PERMISSIONS HAS IN CURRENT DATABASE ...
    (microsoft.public.sqlserver.security)
  • Re: copy permissions from one user to another?
    ... THIS STORED PROCEDURE GENERATES COMMANDS ... -- ADD USER TO SERVER ... -- CREATE TABLE TO HOLD LIST OF USERS IN CURRENT DATABASE ... -- GRANT USER ACCESS TO SERVER ROLES ...
    (microsoft.public.sqlserver.security)