Simon Josefsson | 7 Oct 2001 13:57

IMAP, UseNet and UTF-8 Newsgroups names


In USEFOR it seems as if Newsgroups: names are to be UTF-8 in the
future.  With a little tweaking, this works fine (over NNTP) in the
client I use, but I wonder how it will integrate with IMAP.

First, is it the intention that reading UseNet via IMAP is officially
supported?  It is mentioned in the draft (#news) but everything is
very RFC 822 centric.  (Which btw should probably be updated to 2822.)

Is the UseNet-via-IMAP widely used at all?  Maybe noone cares if UTF-8
newsgroups aren't accessible via IMAP?

The immediate problems that comes to mind is unencoded 8-bit
characters in headers. UTF-8 newsgroups breaks this.  Recommending
that Newsgroups: should be RFC 2047 encoded during IMAP transport may
be one solution, but it has some issues (which charset? Always UTF-8
or just any that matches? etc).

Mark Crispin | 7 Oct 2001 14:40

re: IMAP, UseNet and UTF-8 Newsgroups names


On Sun, 07 Oct 2001 13:57:11 +0200, Simon Josefsson wrote:
> First, is it the intention that reading UseNet via IMAP is officially
> supported?

Yes.

> It is mentioned in the draft (#news) but everything is
> very RFC 822 centric.  (Which btw should probably be updated to 2822.)

I've just made the update.

> Is the UseNet-via-IMAP widely used at all?

Yes, it's used.

> Maybe noone cares if UTF-8
> newsgroups aren't accessible via IMAP?

People do care.

> The immediate problems that comes to mind is unencoded 8-bit
> characters in headers. UTF-8 newsgroups breaks this.  Recommending
> that Newsgroups: should be RFC 2047 encoded during IMAP transport may
> be one solution, but it has some issues (which charset? Always UTF-8
> or just any that matches? etc).

My intention has always been that "all the 8-bit world is UTF-8", and that
8-bit characters in mailbox names and headers are unambiguously UTF-8.

(Continue reading)

Simon Josefsson | 7 Oct 2001 15:20

Re: IMAP, UseNet and UTF-8 Newsgroups names


Mark Crispin <MRC <at> CAC.Washington.EDU> writes:

>> Is the UseNet-via-IMAP widely used at all?
>
> Yes, it's used.

Ok, good. Maybe a reference to the USEFOR message syntax is
appropriate, since news doesn't use RFC 822.

Out of curiosity, can you post to news via IMAP as well?  Is the
distribution based on the APPEND mailbox name, or what is in the
Newsgroups field?

>> The immediate problems that comes to mind is unencoded 8-bit
>> characters in headers. UTF-8 newsgroups breaks this.  Recommending
>> that Newsgroups: should be RFC 2047 encoded during IMAP transport may
>> be one solution, but it has some issues (which charset? Always UTF-8
>> or just any that matches? etc).
>
> My intention has always been that "all the 8-bit world is UTF-8", and that
> 8-bit characters in mailbox names and headers are unambiguously UTF-8.

I think this should be clarified in the draft, right now the draft
references only RFC 822 (+ MIME) which clearly forbids 8-bit
characters in headers.  UTF-8 is not mentioned except for mailbox
names.  Adding a references for message syntax to USEFOR would solve
this (but would require a careful look through the document to see if
something else needs to be updated as well).

(Continue reading)

Mark Crispin | 10 Oct 2001 07:38

Re: IMAP, UseNet and UTF-8 Newsgroups names


On Sun, 07 Oct 2001 15:20:06 +0200, Simon Josefsson wrote:
> Ok, good. Maybe a reference to the USEFOR message syntax is
> appropriate, since news doesn't use RFC 822.

I'm one of those people who believe that the notion that news doesn't use RFC
[2]822 is fundamentally wrong-headed, and as I result I don't accept it.  2822
is the message format syntax.  News is a type of message.

I know that it is possible to argue this point forever.  If you disagree with
me, let's agree to disagree and not waste time arguing about it.

> Out of curiosity, can you post to news via IMAP as well?

There is no posting in IMAP.  IMAP clients use SMTP and NNTP for mail and news
posting, respectively.

> > My intention has always been that "all the 8-bit world is UTF-8", and that
> > 8-bit characters in mailbox names and headers are unambiguously UTF-8.
> I think this should be clarified in the draft, right now the draft
> references only RFC 822 (+ MIME) which clearly forbids 8-bit
> characters in headers.

The prohibition is quite intentional; people should not have the idea that
they have license to do this in IMAP.

> RFC822.HEADER, SEARCH HEADER etc all needs to be described as being
> UTF-8.

Someday, RFC822.HEADER may return a UTF-8 string.  But not now; and for the
(Continue reading)

Alexey Melnikov | 10 Oct 2001 08:22

Question about existing implementations of ACL (RFC 2086)


Hi,

[I am trying to collect some implementation feedback before I submit new
revision of ACL draft.]

Anybody implemented groups in IMAP ACL? What naming schemes different
implementations use?
Any special prefix?

Alexey Melnikov
__________________________________________
R & D, ACI Worldwide (formerly MessagingDirect Ltd.)
phone 780.424.4922 x357

I speak for myself only, not for my employer.
__________________________________________

Alexey Melnikov | 10 Oct 2001 08:47

More questions about IMAP ACL


I have couple generic questions about IMAP ACL:

1). Should the document be fully backward compatible with RFC 2086? (for
example, dropping identifiers starting with dash will break
compatibility).
2). Should the document extend functionality or just fix RFC 2086?

Alexey Melnikov
__________________________________________
R & D, ACI Worldwide (formerly MessagingDirect Ltd.)
phone 780.424.4922 x357

I speak for myself only, not for my employer.
__________________________________________

Mark Crispin | 10 Oct 2001 09:49

re: More questions about IMAP ACL


On Wed, 10 Oct 2001 00:47:43 -0600, Alexey Melnikov wrote:
> 1). Should the document be fully backward compatible with RFC 2086?

Try, but don't allow that to keep you from doing the right thing when
necessary.

> 2). Should the document extend functionality or just fix RFC 2086?

It should extend functionality.

Lawrence Greenfield | 10 Oct 2001 18:38
Picon

Re: Question about existing implementations of ACL (RFC 2086)


   Date: Wed, 10 Oct 2001 00:22:38 -0600
   From: Alexey Melnikov <mel <at> messagingdirect.com>
   Organization: ACI WorldWide / MessagingDirect

   [I am trying to collect some implementation feedback before I submit new
   revision of ACL draft.]

   Anybody implemented groups in IMAP ACL? What naming schemes different
   implementations use?

We implement groups with several different authorization modules:

. The one we use on campus is AFS PTS groups, which don't collide with
  our user identifiers.  (All PTS groups have ":" in them, all PTS
  users don't.)

. The Unix /etc/group authorization module prefixes all Unix groups
  with "group:".

. There's a krb authorization module that allows for wildcards (ie,
  "* <at> CS.CMU.EDU" or "leg.* <at> ANDREW.CMU.EDU") which work like groups.

Larry

Alexey Melnikov | 10 Oct 2001 20:01

Re: More questions about IMAP ACL


Alexey Melnikov wrote:

> I have couple generic questions about IMAP ACL:
>
> 1). Should the document be fully backward compatible with RFC 2086? (for
> example, dropping identifiers starting with dash will break
> compatibility).

It just came to me, that adding new rights is not backward compatible
change: some servers don't support storing of arbitrary rights. Also old
servers will use "d" flag for STORE \Deleted, EXPUNGE and DELETE. This will
be a problem for clients. Should we define a new capability name?

> 2). Should the document extend functionality or just fix RFC 2086?

Alexey Melnikov
__________________________________________
R & D, ACI Worldwide (formerly MessagingDirect Ltd.)
phone 780.424.4922 x357

I speak for myself only, not for my employer.
__________________________________________

Mark Crispin | 10 Oct 2001 20:04

Re: More questions about IMAP ACL


On Wed, 10 Oct 2001 12:01:58 -0600, Alexey Melnikov wrote:
> Should we define a new capability name?

I say yes.  Let's leave ACL as RFC 2086, and define ACLPLUS.


Gmane