Here's a draft of my slightly edited notes from the Sieve BOF last week.
Comments about what I missed or misrepresented are welcome; post 'em here.
Barry
Notes
-----------------------------------------------
Matt says to kick Ned's ass when they see him, because he's late.
Matt reviews mailing list addresses, etc.
Matt reviews agenda.
- Don't want working group.
- Have draft charter, just in case.
- Want to get consensus at end of BOF about WG.
Tim explains syntax briefly.
Tim reviews open issues.
- Vacation in base spec or extension?
reason for base: it's important -- to some, the very reason for having Sieve
reason for extension: it's a long section; it provides a prototype for
writing extensions
consensus: do it as an extension
- Action interaction
e.g. reject vs. keep/fileinto
Proposal:
discard is compatible with everything -- just cancels the default keep
reject is incompatible with all but discard
keep == fileinto "inbox"
Discussion:
add to spec: extensions should discuss incompatible actions
discussion about extensions having to refer to other extensions, easy
to get out of hand.
discuss on mailing list for resolution
- Site limits on actions (number of forward addresses, for instance)
Proposal:
multiple reject MUST NOT
at least one fileinto/keep MUST
at least one redirect MUST
Discussion
redirect: no, remove
exceeding limit causes a script error
- Require must be at beginning of script
- User-specified charset on outgoing msg
Discussion:
text may be MIME object?
OK... put in switch: text or MIME
- Text will be added talking about errors in script
Other Discussion:
- Discussion about adding headers in redirect
Want to add loop detection and received headers
What about resent headers?
Some say it's not user-initiated.
Some consider that it *is* user-initiated.
Consensus... Eliminate text about not adding headers
- Discussion about discarding messages, what to do with DSNs
Take to mailing list
- Vacation extension
Proposal:
Vacation incompatible with reject & vacation.
Does not cancel implicit keep.
Remembering who was responded to is not just per address, but per
address per vacation command.
If script changes, implementations MAY reset response memory.
Also:
Add req't for impl dependent list of addresses that vacation never
replies to, List MUST include POSTMASTER, SHOULD include others.
Discussion:
Remove specific addresses, replace with general advice, refer to RFCs.
Request to be very explicit about this, to avoid loops & annoyances
calendar in vacation? -- out of scope
Matt:
Proposes getting revision out in 3 weeks, for IETF last call (4 weeks).
Includes base & vacation. Call for WG... unanimous: no.