Re: freedomization task list [was: Re: Dangerous precedent being set - possible serious violation of the GPL]
Caspian <caspian@twu.net> writes:
> I've thought about this issue a bit.
>
> Would it be possible, when reading in the mailbox, to simply scan the
> thing for headers-- not keeping the entire mailbox in memory? When someone
> wants to read a given message, it would open the message by scanning
> forwards to the message, and reading. Or would even this be horribly slow?
>
> It might break compatibility, but I think a really neat feature would be a
> library that would allow people reading large files (e.g. mail spools) to
> actually start reading -right- at the disk sector that contains what they
> want...bypassing traditional file reading operations altogether (e.g. open
> the file, scan to where you want, start reading...)... this way, the
> program could maintain an index-- okay, message 1 starts at raw address X,
> message 2 at raw address Y, and so on. Would this be do-able? It'd be a
> nice kluge, certainly, but I'm afraid I know very little ab out how
> filesystems work...
Actually, this is exactly what libmailbox does. Oh, you probably
haven't heard of libmailbox. :) It's part of the as-of-yet unreleased
(I keep saying that) GNU mailutils package. See
http://www.gnu.org/software/devel.html#CVS for info on getting it out
of CVS. Another neat feature of libmailbox is that it works with
numerous mailbox types (mbox and maildir already partially
implemented, pop3 and imap4 in the works).
--
Jakob 'sparky' Kaivo - jkaivo@ndn.net - http://jakob.kaivo.net/
Reply to: