[Tickets #8936] Clean up date/arrival/sequence sorting
bugs at horde.org
bugs at horde.org
Fri Mar 26 22:14:29 UTC 2010
DO NOT REPLY TO THIS MESSAGE. THIS EMAIL ADDRESS IS NOT MONITORED.
Ticket URL: http://bugs.horde.org/ticket/8936
------------------------------------------------------------------------------
Ticket | 8936
Created By | Michael Slusarz <slusarz at horde.org>
Summary | Clean up date/arrival/sequence sorting
Queue | IMP
Version | Git master
Type | Enhancement
State | Accepted
Priority | 1. Low
Milestone | 5.0
Patch |
Owners |
------------------------------------------------------------------------------
Michael Slusarz <slusarz at horde.org> (2010-03-26 18:14) wrote:
Date sort = date in message
Arrival sort = internal imap date of message
sequence sort = straight numeric sort of sequence numbers (order in
which messages were added to the mailbox)
Sequence sort needs to be default because it is the only reasonable
sort that can be done without parsing headers across all IMAP servers
Arrival sort in IMP 4.x referred to sequence sort - arrival sort was
never defined in RFC 3501 but it makes sense that it is the arrival
order into the mailbox. Additionally, arrival/sequence sort (at least
in IMP) had the sequence numbers which clearly indicates the sort was
being done on sequence numbers (not that this means anything to the
average user). Then RFC 5256 goes and confuses thing but defining a
SORT ARRIVAL which sorts by internaldate.
In IMP 5, to clean this up, do the following:
1.) Rewrite Horde_Imap_Client to handle sequence sort - obtained doing
a SEARCH (not a SORT on imap servers that support the latter) and then
doing numeric sorting on the client side.
2.) IMP defaults to sequence sort for same reason as in IMP 4.x.
3.) Only display one type of date search in mailbox headings.
Determining whether this does a Header Date or an Internaldate search
is left to a preference, locked by default. By default, resort to
Internaldate (since this tends to be more accurate than the Date
header - see, e.g., Spam headers dated way in the future/past).
4.) Determine how to allow user to resort back to sequence sort. Call
this sort "no sorting'.
More information about the bugs
mailing list