grephistory?

Richard Michael Todd rmtodd at mailhost.ecn.ou.edu
Wed Aug 25 21:11:42 UTC 1999


In message <37C41018.221E66BB at ubc.ca>you write:
>
>That pretty wells describes my situation. Fortunately I'm still running
>in slave mode (in preparation to cutting over) so I don't have any
>duplicate hashes (yet). 
>
>If tagged-hash is so broken (ie. useless) why is it still an option?

Well, in my experience it isn't broken.  There are issues with the
.pag file not handling overflow gracefully, causing duplicates, but
that only happens when one is starting up news for the first time and
the history file hasn't reached a steady-state size, but once the
history file and .pag file have reached their steady-state sizes, this
doesn't happen.  Admittedly this is a problem, but it's hard to fix
without having some way of knowing what the target steady-state
history size will be, which is difficult since history size varies so widely
depending on local spool size and expiration policy. 

More to the point, tagged hash takes only about half as much memory for expire 
and innd to run as the non-tagged-hash version.  Simply put, without tagged
hash, a lot of sites, including this one, would not have the memory to run 
INN.  


More information about the inn-workers mailing list