Cyrus Daboo | 9 Dec 20:00 2002

managesieve comments


Some comments on the managesieve -04 draft based on client implementation 
experience:

Comments on Manage SIEVE

- According to the current IANA registry port 2000 is already assigned to 
'callbook'. Looks like we have to pick another default port?

- STARTTLS command description still has text stating that server must send 
capability items.

- 'sieve-name' item in formal syntax is just defined as 'string'. Why not 
just use 'string' instead? If you want to impose the semantic script name 
requirement, then the 'sieve-name' item should have a comment with a 
reference to section 1.7.

- 'command-authenticate' is missing the mandatory CRLF at the end. The 
correct syntax should be:

    command-authenticate  = "AUTHENTICATE" SP auth-type [SP string]
                            *(CRLF string) CRLF

- 'extension-data' in 'resp-code-ext' is not defined in this document. It 
is in ACAP, but there is no comment about dependency on ACAP formal syntax 
for undefined items here. Either add 'extension-data' in this doc or add 
reference to ACAP in the formal syntax introduction paragraph.

- There is no support for disconnected/multi-access behaviour. In 
particular, a client has to always upload a script before making it active 
(Continue reading)

Tony Hansen | 9 Dec 22:54 2002
Picon

Re: managesieve comments


Cyrus Daboo wrote:
> 
> Some comments on the managesieve -04 draft based on client 
> implementation experience:
> 
> Comments on Manage SIEVE
> 
> - According to the current IANA registry port 2000 is already assigned 
> to 'callbook'. Looks like we have to pick another default port?

We don't pick ports -- IANA assigns them.

The port should be listed as TBA (to be assigned), and a note put in the 
IANA Considerations section indicating that they need to register a port 
for this purpose.

	Tony Hansen
	tony <at> att.com

Marc Mutz | 9 Dec 23:02 2002
Picon

Re: managesieve comments

On Monday 09 December 2002 20:00, Cyrus Daboo wrote:
> Some comments on the managesieve -04 draft
<snip>

Oh, you have a copy? Can you send it?

Marc

--

-- 
They that can give up essential liberty to obtain a little temporary
safety deserve neither liberty nor safety.      -- Benjamin Franklin
Cyrus Daboo | 10 Dec 00:26 2002

Re: managesieve comments

Hi Marc,

--On Monday, December 9, 2002 11:02 PM +0100 Marc Mutz <mutz <at> kde.org> wrote:

|> Some comments on the managesieve -04 draft
| <snip>
|
| Oh, you have a copy? Can you send it?

Try:

<ftp://ftp.ietf.org/internet-drafts/draft-martin-managesieve-04.txt>

--

-- 
Cyrus Daboo
Lyndon Nerenberg | 10 Dec 00:42 2002
Picon

Re: managesieve comments


>- There is no support for disconnected/multi-access behaviour. In 
>particular, a client has to always upload a script before making it active 
>as it has no way of knowing whether the one it uploaded in a previous 
>session has been changed by some other client. This will be an issue for 
>large scripts. The fix for this would be to have some kind of modtime or 
>uid stamp for each script - this is probably not necessary right now, but 
>may become important later.

Can we just let the server return the MD5 digest of the script?

--lyndon

Cyrus Daboo | 10 Dec 01:08 2002

Re: managesieve comments


Hi Lyndon,

--On Monday, December 9, 2002 4:42 PM -0700 Lyndon Nerenberg 
<lyndon <at> orthanc.ab.ca> wrote:

|> - There is no support for disconnected/multi-access behaviour. In
|> particular, a client has to always upload a script before making it
|> active  as it has no way of knowing whether the one it uploaded in a
|> previous  session has been changed by some other client. This will be an
|> issue for  large scripts. The fix for this would be to have some kind of
|> modtime or  uid stamp for each script - this is probably not necessary
|> right now, but  may become important later.
|
| Can we just let the server return the MD5 digest of the script?

That would work. Are you suggesting a new command to return the MD5 of a 
named script?

I was thinking in IMAP terms of introducing a UID command and have that 
modify existing command behaviour:

UID LISTSCRIPTS - this would return the list of scripts with the UID for 
each script listed in the responses.

UID PUTSCRIPT - this would return the new UID as a response code in the OK 
response if it succeeded.

UID SETACTIVE - the client would provide a UID value in the command, and if 
that did not match the UID for the stored named script, the command would 
(Continue reading)

Andrew Hammond | 16 Dec 18:36 2002

dynamic naming & creation of mailboxes


Hi folks.

I'm looking to create a new sub-folder / day setup for an mailbox that gets
a couple of hundred messages every day.  Then automatically deliver messages
to today's folder.

I've checked the mailing list archives, googled, and read the Sieve RFC.  It
looks to me like there's now way to do this easily with Sieve.

My current solution is a perl script run by a cron job that makes the
sub-folder and writes a new .sieve.

Is there a better way to do this?

Cyrus Daboo | 16 Dec 19:07 2002

Re: dynamic naming & creation of mailboxes


Hi Andrew,

--On Monday, December 16, 2002 12:36 PM -0500 Andrew Hammond 
<andrew.hammond <at> surveysite.com> wrote:

| I'm looking to create a new sub-folder / day setup for an mailbox that
| gets a couple of hundred messages every day.  Then automatically deliver
| messages to today's folder.
|
| I've checked the mailing list archives, googled, and read the Sieve RFC.
| It looks to me like there's now way to do this easily with Sieve.
|
| My current solution is a perl script run by a cron job that makes the
| sub-folder and writes a new .sieve.
|
| Is there a better way to do this?

Why not leave the sieve script set to deliver into a mailbox called 
'sent-active' all the time, then simply have the perl script rename that to 
yesterday's date and recreate the active mailbox again. You then do not 
need to edit the sieve script each day. The only problem with this is a 
possible race condition in which a message delivery to the mailbox is 
attempted between the rename and re-create actions. An alternative might be 
to create yesterday's mailbox and then copy all messages from the active 
one into that and expunge those. That way the active mailbox is around all 
the time avoiding the race. However, that requires a lot more work on the 
IMAP side as you have to select (open) the mailbox to do the copying.

--

-- 
(Continue reading)

Ken Murchison | 16 Dec 19:23 2002

Re: dynamic naming & creation of mailboxes


Cyrus Daboo wrote:
> 
> Hi Andrew,
> 
> --On Monday, December 16, 2002 12:36 PM -0500 Andrew Hammond
> <andrew.hammond <at> surveysite.com> wrote:
> 
> | I'm looking to create a new sub-folder / day setup for an mailbox that
> | gets a couple of hundred messages every day.  Then automatically deliver
> | messages to today's folder.
> |
> | I've checked the mailing list archives, googled, and read the Sieve RFC.
> | It looks to me like there's now way to do this easily with Sieve.
> |
> | My current solution is a perl script run by a cron job that makes the
> | sub-folder and writes a new .sieve.
> |
> | Is there a better way to do this?
> 
> Why not leave the sieve script set to deliver into a mailbox called
> 'sent-active' all the time, then simply have the perl script rename that to
> yesterday's date and recreate the active mailbox again. You then do not
> need to edit the sieve script each day. The only problem with this is a
> possible race condition in which a message delivery to the mailbox is
> attempted between the rename and re-create actions. An alternative might be
> to create yesterday's mailbox and then copy all messages from the active
> one into that and expunge those. That way the active mailbox is around all
> the time avoiding the race. However, that requires a lot more work on the
> IMAP side as you have to select (open) the mailbox to do the copying.
(Continue reading)

Kjetil Torgrim Homme | 17 Dec 00:19 2002
Picon
Picon

Re: dynamic naming & creation of mailboxes


[Andrew Hammond]:
>
>   I'm looking to create a new sub-folder / day setup for an mailbox
>   that gets a couple of hundred messages every day.  Then
>   automatically deliver messages to today's folder.
>   
>   I've checked the mailing list archives, googled, and read the
>   Sieve RFC.  It looks to me like there's now way to do this easily
>   with Sieve.

I'm currently writing a draft for an extension allowing expansion
inside strings.  the main motivation for it is to allow access to back
references in regular expressions.  with that in place (I hope to hack
an implementation during Christmas :-), you could do

  require ["fileinto", "regex", "stringexpand"];

  if header :regex ["date"] ", ([0-3]?[0-9]) ([A-Z][a-z]{2}) ([0-9]{4})" {
        let "year"  "${3}";
        let "month" "${2}";
        let "day"   "${1}";
        fileinto "INBOX.${year}-${month}-${day}";
  }

this is a bit awkward, though, and you are not guaranteed that the
Date header will comply with RFC 2822 grammar.  (most of the bogus
Date headers would probably be found in spam messages, though.)

perhaps the extension should include built-in variable names for time
(Continue reading)


Gmane