ietf-822
[Top] [All Lists]

Re: 8-bit transmission in NNTP

1994-09-12 14:30:08
keld(_at_)dkuug(_dot_)dk (Keld J|rn Simonsen) writes:
Well, news is a store-and-forward system. There is not much difference
between news and mailing list distribution.

In a mailing list distribution, a message goes to a central point, and
then is delivered to each subscriber using a point-to-point system.
The path (or small set of paths) by which a subscriber receives
mailing lists is fixed.

News, on the other hand, is a flooding system.  Each site offers to
send each article to each of its neighbors.  There are many cycles in
the distribution path, each site only accepts those articles it hasn't
previously gotten from somewhere else. The paths by which subscribers
receive news are highly variable and timing dependent.

If there is a transport agent which munges the contents of messages
which pass through it, the consequences differ greatly between the two
systems.  In mail, it only affects those sites which receive mail
through that agent.  The munging is consistent, and the sites can
avoid it by arranging to receive mail through some other path.

In news, there is not necessarily a "downstream".  If an article is
munged while going through one path, there is usually no limit on
where the munged copy will go--it will depend on timing.

If 8bit-->7bit transfer-encoders are widely deployed in news, then
sites in 8bit "enclaves" will see transfer-encoded messages which have
passed out of the enclave and back in.  If anyone ever deploys a
transfer-encoder with a bug, that bug will randomly affect everyone
until all instances of the bug are eradicated from the planet.

Header munging used to continuously cause problems in news, that was only
fixed when the news transport authors (particulary Henry Spencer)
instituted the "thou shalt not rewrite headers during transport" rule
and started dropping articles not conforming to the spec on the floor.

-- 
_.John G. Myers         Internet: jgm+(_at_)CMU(_dot_)EDU
                        LoseNet:  ...!seismo!ihnp4!wiscvm.wisc.edu!give!up

<Prev in Thread] Current Thread [Next in Thread>