Path: utzoo!utstat!helios.physics.utoronto.ca!news-server.csri.toronto.edu!math.lsa.umich.edu!zaphod.mps.ohio-state.edu!usc!orion.oac.uci.edu!cedman From: cedman@golem.ps.uci.edu (Carl Edman) Newsgroups: news.software.nntp Subject: Re: fix for nntp 1.5.10 / GNUS problem Message-ID: Date: 23 Nov 90 00:51:47 GMT References: <1990Nov20.221152.10825@csn.org> Organization: University of California, Irvine, USA. Lines: 45 Nntp-Posting-Host: lynx.ps.uci.edu In-reply-to: mberkley@sanjuan.uvic.ca's message of 21 Nov 90 16:41:10 GMT In article mberkley@sanjuan.uvic.ca (Mike Berkley - 721-8752) writes: > On 21 Nov 90 03:21:57 GMT, cedman@golem.ps.uci.edu (Carl Edman) said: CE> About the hack to fix it. I usually post it once a week in CE> gnu.emacs.gnus (that is how often the problem comes up :-), but CE> here it is: CE> 2. Find where TIMEOUT is defined CE> 3. Comment it out CE> Negative consequence: There used to be a timeout for idle NNTP CE> connections of about 2 hours. This is gone. I (in fact this CE> university) can live(and post) with this just fine, but your CE> mileage may vary. But this university cannot live with this fix - our administrator (rightly) exhumed the buggy NNTP 1.5.10 after he found 86 nntpd's running (no timeouts). Vick Khera's fix is much more acceptable: [...] Please ! Not again the same discussion we have every week on gnu.emacs.gnus. The main point against this fix is that it impacts the preformance of NNTP for the users. The consensus was that this probably wasn't quite as bad as setting nntp-maximum-request to 1 in gnus, but still significant. There can also be made another point against this hack. There is a bug in the NNTP software, which is at the core of the problem: The mixing of high-level system-buffered I/O and low-level I/O, and that is simply not a good thing. Removing buffering is simply another kludge to disable the effects of this bug. Turning of the TIMEOUT removes at least one instance of this "cross-programming" (If I may coin the term :-). So, decide what tradeoffs you will make in your installation, but know the advantages and disadvantages of all options. Carl Edman Theorectical Physicist,N.:A physicist whose | Send mail existence is postulated, to make the numbers | to balance but who is never actually observed | cedman@golem.ps.uci.edu in the laboratory. | edmanc@uciph0.ps.uci.edu