Re: Writing a GNOME mail client.

[Warren Young <tangent cyberport com>]

Scott Wimer wrote:
> 
> We've really only got 2 options to us, each Message in a single file,
> or some aggregation method for putting multiple Messages into a single
> file.  Any aggregation method is going to take some additional
> algorithmic complexity to handle.  Maybe, if all the Messages for a
> given day were placed into a single file, that would make them fairly
> easy to retrieve, since we can resolve the storage file from the
> Message ID rather painlessly.

I addressed this in another message, and it's something that may work
better than any other scheme.  Basically, each message is still split up
into parts, but we use an ar(1) or cpio(1) archive to hold the parts. 
There must be libraries we can swipe that make it easy to load these
archives.

Someone also mentioned the overhead this scheme has: actually, it has
very little.  Instead of having to parse a complete message each time,
we can parse only what I call the control file, which is in a
rigidly-defined format, and thus is easy to parse quickly.  It's also
faster than any database mechanism, because simple file lookups would
almost have to be faster than database queries, no?
-- 
= Warren -- http://www.cyberport.com/~tangent/
= ICBM Address: 36.8274040 N, 108.0204086 W, alt. 1714m
= Of all the people I have met, you are certainly one.