jbrown@herron.uucp (Jordan Brown) (12/14/88)
Looks to me like somebody out there is corrupting news batches. I've seen several articles which end "#! rnews" and maybe a bit more, which looks like an unbatcher got out of sync. This could very well be eating messages. Perhaps somebody counted CRs when generating the batch. Unfortunately, both messages went through almost exactly the same path, so there's no trivial way to point a finger. However, the fact that they did may imply that the damage was done early. The bitnet transition is perhaps a good place to look. Examples below: --- Path: herron.uucp!jato!elroy!ames!haven!uflorida!gatech!psuvax1!psuvm.bitnet!cunyvm!ndsuvm1!ndsuvax!ncreed From: ncreed@ndsuvax.UUCP (W. Reed) Newsgroups: news.admin Subject: Need uucp / News connection in Twin Cities (Minnisota) Message-ID: <1947@ndsuvax.UUCP> Date: 9 Dec 88 02:46:41 GMT Distribution: usa Organization: North Dakota State University Fargo, ND Lines: 13 [ message text deleted] #! rnews --- Path: herron.uucp!jato!elroy!ames!haven!uflorida!gatech!psuvax1!psuvm.bitnet!cunyvm!ndsuvm1!ndsuvax!numork From: numork@ndsuvax.UUCP (James Mork) Newsgroups: soc.culture.jewish,news.misc,news.sysadmin Subject: Re: Anti-Semite witch hunt Message-ID: <1948@ndsuvax.UUCP> Date: 9 Dec 88 03:36:49 GMT References: <49196@pyramid.pyramid.com> <401@hropus.UUCP> <Dec.6.17.28.54.1988.7817@topaz.rutgers.edu> <1017@lzfme.att.com> Reply-To: numork@ndsuvax.UUCP (James Mork) Organization: North Dakota State University, Fargo Lines: 13 [ message text deleted ] #! rnews 757 Path: ---
dhp@att.ATT.COM (Price) (12/16/88)
It seems that this occurs when news 2.11 runs out of space in the news filesystem while unpacking batches. Intermediate writes may or may not make it to the disk, causing the byte counts in the #rnews line to get out of sync. The solution would be to have the code notice write errors and give up immediately (preferably in some fault tolerant fashion that would preserve the data. -- Douglas H. Price Postmaster, att att!dhp