Latent bug in the new tradindexed code?

Katsuhiro Kondou Katsuhiro_Kondou at isc.org
Wed Sep 18 05:24:51 UTC 2002


In article <ylwupkvd32.fsf at windlord.stanford.edu>,
	Russ Allbery <rra at stanford.edu> wrote;

} The right way to avoid the copy, if someone really cares, is to provide an
} interface to the overview API that says "here's a range, here's a file
} descriptor, spew overview information from that range down that file
} descriptor."  That'll be the fastest since if the overview information
} happens to all be sequential, you can even use sendfile.

First, I must confess using sendfile for buffindexed, as
noted before in another thread, at nnrpd may not work in
certain situation.  There can be a case that the data block
is already freed by expireover when the data is really read.
So, as to buffindexed it still requires memory allocation
for data retrieval internally(buffindexed.c).

I'd want to avoid further data copying at nnrpd, and this
is why I wrote such API.
-- 
Katsuhiro Kondou


More information about the inn-workers mailing list