K 10
svn:author
V 7
coneill
K 8
svn:date
V 27
1997-08-27T03:49:48.000000Z
K 7
svn:log
V 1243
Whether you use this option is configured with the 'timespool'
option in inn.conf.  It defaults to off.

This changes the spool organization somewhat.  It makes the normal
hierarchy a shadow of a time based hierarchy.  The time based hierarchy
is two levels.  The first level is bits 20-13 of the time that the article
was received in hex, and the second level is bits 13-6 of the same.

The article name name is of the format tttttttt-ssss, where tttttttt is the
time the article was recieved in time_t format, and ssss is a sequence
number to distinguish article that were recieved in the same second.

The primary advantage of this is for feed-only machines, where you
only access articles in the order written.  In that case it gives
extremely good locality of reference.   This also helps to spread articles
evenly across the time spool.  For reader machines this can be a help
if you aren't i/o bound, except that open takes forever.  Writing the
article into the time spool first and then letting crosspost handle all the
real article links can be a performance win depending on the circumstances.
I would expect that people running this on feed-only machines would
just turn off crossposts altogether, using only the time based spool.

END
