On Monday 26 June 2006 16:20, Daniel W wrote:
> Brian G. Peterson wrote:
> > On Monday 26 June 2006 13:54, Daniel W wrote:
> >> Just wondering if you heard of Mairix?
> >> This is an IMAP folder index searcher. Seems like a good idea and
> >> gives some fantastic (FAST) search capabilities which might be
> >> really attractive if used within the squirrelmail search page
> >> (although I know you are planning a more advanced version?).
> > The SM DEVEL code contains a fully RFC compliant advanced search
> > page. If you feel that the Advanced Search capabilities in the Devel
> > code is not sufficient, please submit patches.
> Yes sorry I do need to check out this facility. However I think this is
> still based on server-side seraching which requires physical reading of
> the actual message data (eg very expensive with mbox).
Dovecot supports mbox mail folders and a separate (very fast) index to
avoid this problem. All the server-side indexes that I'm aware of index
Header data only, since most searches are on Header data. Some clients
have indexes that include <body> data, but that's pretty rare. Body
searches are usually quite slow no matter what.
Mairix seems to work by keeping a separate index of all words (including
in the body) and then creating a virtual Maildir file for the client to
access. With most IMAP servers, this won't work. Mairix seems designed
specifically for local clients that can reed mail folders from the
filesystem in multiple formats (mbox, Maildir, or MH) like Sylpheed,mutt,
> >> This might be better as a plugin but I imagine it would replace the
> >> current search facility if enabled by the admin.
> >> http://freshmeat.net/projects/mairix/
> > Definitely best as a plugin. Some IMAP servers already keep very
> > fast indexes separately from the folders (Cyrus, Dovecot come to
> > mind), so on these imap servers, searches are already basically
> > instantaneous for anything other than a full-text search.
> Aren't these just things like mail sender, recipients, date?
All Headers, yes.
> What about subject, body? Those (along with sender) are the most likely
> places people are going to search.
Subject is a Header, and is indexed by all the IMAP servers that I'm aware
of that use indexing, Body isn't a header, and isn't indexed. Size and
attachment names are also headers, iirc.
> Mairix looks likely to be able to do full body indexing.
Yes Mairix indexes all 'words' in the body, but then manipulates the
filesystems to create a virtual Maildir folder. Most IMAP servers won't
be able to access this, as the client won't have it subscribed (and it
may not be in the correct place in the filesystem, but that should be
able to be worked around.
I'm a little unclear what you feel the performance advantages of this
would be over other server-side searching, for anything other than the
body of the message. For managing mbox files, I think the performance
would be worse as Mairix needs to create a virtual search results Maildir
and then copy messages out of the mbox file(s) and into the search
results Maildir that it creates.
So, clearly this would only be workable as a special plugin to
Squirrelmail, and I'm not sure it's even technically feasible in most
IMAP server configurations.