2.5 wish list (was Re: pre-2.4 (was Re: buffindexed jumbo patch) )
Kai Henningsen
kaih at khms.westfalen.de
Sun Jan 12 07:33:00 UTC 2003
rra at stanford.edu (Russ Allbery) wrote on 11.01.03 in <ylvg0vujmt.fsf at windlord.stanford.edu>:
> The only thing I was worried about with this was the speed hit and
> complexity of parsing the Xref header out of the article. (It would be
> *so* nice to store newsgroup and article number information with the
> article in a machine-readable form rather than only as an Xref header in
> the article; the parsing would be so much easier.)
As an aside, that's approximately what I'm doing in my current project
(which is a combined MUA/newsreader with its own messagebase). Message-on-
disk-format is approximately <length of raw msg> <raw msg> <machine
readable msg attributes>. Those attributes, though, mainly cover what can
*not* be reconstructed from the raw message. (Both those and some from the
raw message get put into the db files for fast lookup - most importantly,
threading just needs to look at the db - but the point for me is I can
throw away the db (it's a pure cache) and completely reconstruct it from
the message files, and the raw message part of those files never gets
rewritten.) Oh, and I'm going to use md5's of the message ids - shorter
and completely collision free on half a million ids I could lay my hands
on for a quick test.
MfG Kai
More information about the inn-workers
mailing list