[net.news] Revised Version of News Batching

pag@sri-unix (11/27/82)

Not too long ago John Levine (ima!johnl) submitted the sources to his
news batching software to net.sources.  I have since modified them slightly
to be more general, and have found batching to be an unqualified success
in getting rid of uucp problems.  In order to encourage universal use
of this system, I am submitting my revised version to net.sources.  I
heartily recommend as many sites as possible to install it.

--peter gross
ucbvax!hplabs!hao!pag
seismo!hao!pag

bstempleton (11/28/82)

People who install batching systems should make sure they don't slow news
down too much.  If somebody installs news batching with a 1 hour delay along
all the sites in a 14 site chain, we add a significant delay to news.

The only valid way to solve this is to fix uucp and uux to not do as much
stuff with each file.  You get the same result, only less kludged and with
less delay.

Shorter delays mean less garbage articles because nobody posts 300 explanations
of what foobar means.

mark (11/28/82)

Batching can make a significant performance difference if you run
UUCP.  I wish someone would dig into UUCP and figure out why it's
so inefficient and fix it.

We batch news on cbosgd for a different reason: we don't want the
system loaded down during the day when people are working.  We do
flush out news every 2 hours except during the day.  But we aren't
in the backbone.

Those sites that are in the backbone should be encouraged to either
not batch news, or to flush the queues as often as possible (every
hour is probably perfectly adequate).  Those sites on spurs should
take into account the needs of the sites they feed, but batching to
improve performance or cut down on phone bills is probably a very
reasonable thing to do, provided the delay inserted is not more
than several hours.

	Mark

wagner (11/28/82)

I agree with Brad.  Perhaps the same amount of effort spent
finding out why UUCP was slow would yield a faster, more
stream-lined UUCP.  Adding more layers of complication in the
form of batching seems like a step in the wrong direction.
(note that I am not volunteering, since I have neither time
nor expertise to do either)

Michael Wagner, UTCS