mailing list archives

Has anyone tried how badly clients would break if FETCH didn't send
replies ordered by the sequence number? So for example:

1 fetch 1:* body.peek[]
* 3 FETCH (BODY.PEEK[] "")
* 1 FETCH (BODY.PEEK[] "")
* 2 FETCH (BODY.PEEK[] "")
1 OK

I guess this is a bad idea.

-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 196 bytes
Desc: This is a digitally signed message part
URL: <http://mailman13.u.washington.edu/pipermail/imap-protocol/attachments/20070723/2024e1b4/attachment.sig>

On Mon, 23 Jul 2007, Timo Sirainen wrote:
> Has anyone tried how badly clients would break if FETCH didn't send
> replies ordered by the sequence number? So for example:
>
> 1 fetch 1:* body.peek[]
> * 3 FETCH (BODY.PEEK[] "")
> * 1 FETCH (BODY.PEEK[] "")
> * 2 FETCH (BODY.PEEK[] "")
> 1 OK

No compliant client will break with this.

Nor, for that matter, will a compliant client break if the response for a 
message is broken into multiple FETCH responses.  Under certain 
circumstances, UW imapd will do that.

Proper handling of this is a requirement of IMAP clients.

-- Mark --

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

Timo Sirainen writes:
> Has anyone tried how badly clients would break if FETCH didn't send 
> replies ordered by the sequence number?

Yes, and no breakage. Clients break on many things they should handle 
(try sending lowercase OK/NO/BAD ;) but that one seems fine.

(We changed to strict increasing order later because of testing constraints.)

Arnt

On Mon, 23 Jul 2007, Arnt Gulbrandsen wrote:
> Yes, and no breakage. Clients break on many things they should handle (try 
> sending lowercase OK/NO/BAD ;) but that one seems fine.

Which clients break on that?  They and their vendors should be publicly 
humiliated.

-- Mark --

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

Mark Crispin writes:
> On Mon, 23 Jul 2007, Arnt Gulbrandsen wrote:
>> Yes, and no breakage. Clients break on many things they should handle 
>> (try sending lowercase OK/NO/BAD ;) but that one seems fine.
>
> Which clients break on that?

I don't remember. Some bug tracking system may remember, but I've left 
the relevant company and no longer have access to that information.

> They and their vendors should be publicly humiliated.

It went like this:

A fool made OK/NO/BAD lowercase sometimes and got it released. One day 
later the first breaking client showed up. The fool saw that it was an 
open source client (mutt, as it happens), and submitted a very simple 
patch, and the patch was accepted at once. Over the next week, n more 
clients showed up.

I wish uw, cyrus or courier would send lowercase IMAP keywords to flush 
these bugs out.

Signed,
a fool

On Mon, 23 Jul 2007, Arnt Gulbrandsen wrote:
> A fool made OK/NO/BAD lowercase sometimes and got it released. One day later 
> the first breaking client showed up. The fool saw that it was an open source 
> client (mutt, as it happens), and submitted a very simple patch, and the 
> patch was accepted at once. Over the next week, n more clients showed up.

You should have stuck to your guns.  Had we known, we would have backed 
you up.

> I wish uw, cyrus or courier would send lowercase IMAP keywords to flush these 
> bugs out.

I would be willing to do this, but not at the cost of having the author of 
Courier use this as FUD against my server.

-- Mark --

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

Mark Crispin writes:
> On Mon, 23 Jul 2007, Arnt Gulbrandsen wrote:
>> I wish uw, cyrus or courier would send lowercase IMAP keywords to 
>> flush these bugs out.
>
> I would be willing to do this, but not at the cost of having the 
> author of Courier use this as FUD against my server.

Similar considerations prevailed at the time. *shrug*

Arnt