X-Status flag, reconstruct and a suggestion

[Rob Siemborski]

On Fri, 3 Oct 2003, Diego Rivera wrote:

> Someone on this list replied that it was a good idea, and a patch could
> be warranted for this functionality.  I'd like to ask what the status of
> that is, and how I may help.

I think I meant that "if you write it, I'll look at it and strongly
consider including it."  Not "I'll write that."

> Secondly, it seems to me that a drawback from cyrus not keeping the
> X-Status flags (for efficiency, and I understand this) is that if the
> databases are corrupted, the state of a user's mailbox is lost
> completely and all mails are re-marked as "unread".
>
> So - would there be any negative impact to adding to Cyrus-IMAP the
> ability to mirror the bits in the index database for these status flags
> in an X-Status header, for added robustnes (at the potential cost of a
> performance hit when viewing e-mail).

Writing out X-Status flags would be a tremendous performance penalty on
the Cyrus Server.  It requires rewriting the message files for every flag
change *and* removing the X-Status header when transmitting the message
via imap, since for a given IMAP UID the message is immutable, so headers
cannot appear and dissapear.

> Would this be viable (as a default-off option even)?  Who would I
> coordinate with to help in this regard?

Unlike the reconstruct patch, which is helpful for migration, I don't
think we would ever take something that modified the contents of message
files.

-Rob

-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
Rob Siemborski * Andrew Systems Group * Cyert Hall 207 * 412-268-7456
Research Systems Programmer * /usr/contributed Gatekeeper