Path: utzoo!utstat!news-server.csri.toronto.edu!math.lsa.umich.edu!zaphod.mps.ohio-state.edu!swrinde!ucsd!ucbvax!ucbarpa.Berkeley.EDU!fair From: fair@ucbarpa.Berkeley.EDU (Erik E. Fair) Newsgroups: news.software.nntp Subject: Re: Inbound news is garbled Message-ID: <38821@ucbvax.BERKELEY.EDU> Date: 26 Sep 90 11:39:51 GMT References: <26339@cs.yale.edu> Sender: usenet@ucbvax.BERKELEY.EDU Organization: USENET Protocol Police, Western Gateway Division Lines: 24 You all should be aware that there is a fundamental assumption in the NNTP 1.5 nntpxmit program (which was safe to make at the time that I wrote it): failures are transient. If there is a non-transient failure in the remote, the sending system will pound the remote with all the articles that cause the failure until the articles are either: 1. successfully accepted by the remote 2. expired on the sending system The previous behavior was that failures were logged, but the article was not reoffered, because there was no requeue code in nntpxmit. This meant that unless you have redundant feeds, you stand to lose some news to transient failures. I chose to try and reduce this possibility, and all testing done prior to release indicated that all failures seen by nntpxmit from the remote were indeed transient. So why did I bring this up? Well, the failure counts could be caused by one article being repeatedly failed by the remote, until it is accepted later on... Erik E. Fair ucbvax!fair fair@ucbarpa.berkeley.edu