The official _ _ _ __ _ __ | |_ _ __ ___ __ _ ___| |__ ___ | '_ \| '_ \| __| '_ \ / __/ _` |/ __| '_ \ / _ \ | | | | | | | |_| |_) | (_| (_| | (__| | | | __/ |_| |_|_| |_|\__| .__/ \___\__,_|\___|_| |_|\___| |_| __ / _| __ _ __ _ | |_ / _` |/ _` | | _| (_| | (_| | |_| \__,_|\__, | |_| NNTPCACHE FREQUENTLY ASKED QUESTIONS AND ANSWERS Peter Tonoli <anarchie@suburbia.net> This is the list of frequently asked questions, and the answers to those questions for nntpcache, the caching NNRP server daemon. ========================================================================= Index Section 1. Introduction and General Information Q1.1 What is nntpcache? Q1.2 What operating systems does nntpcache support? Q1.3 Where can I find nntpcache? / Where is the latest version? Q1.4 Can I use nntpcache freely or do I have to license it? Section 2. Frequently encountered problems and solutions Q2.1 My system chokes on all that damned mmap()ing?!. What can I do? Q2.2 Netscape-news kills itself when reading news from nntpcache-0.88.3! Q2.3 Expiring with nntpcached -e doesn't work Q2.4 Help! The size of our cache keeps growing! What should I do about expiration? Q2.5 nntpcache spits out : 'post failed From: address not in Internet syntax'. What does that mean? Q2.6 Why does nntpcache-update take so long to load at initialisation? Section 3. Support and Resources Q3.1 Is there a nntpcache mailinglist? Q3.2 How do I unsubscribe from the mailinglist? Q3.3 Is there an archive of the nntpcache mailinglist? Q3.4 Is there an official nntpcache home page? Q3.5 Who should I contact if I can't get nntpcache to work / I've found a bug in nntpcache, what should I do? Section 4. Configuration of nntpcache Q4.1 What are all those entries in the servers configuration file? Q4.2 How do I specify the "Xref:full" part in the "overviewFmt" line in the config file? Q4.3 I've entered a server but nntpcache doesn't seem to want to know about it. Q4.4 I want to set the nntpcache-virtual-server-size record, what are some public news-servers I can add to my servers file? -------------------------------------------------------------------------- Question 1.1. What is nntpcache? Nntpcache (efficiently) executes on the localhost pretending to be an NNRP news reading server. In fact, what it does is pass certain NNTP commands through to real (remote and possibly local) news-servers based on various pattern matching rules. nntpcache then takes the output from those servers and caches & indexes it in funky ways (much specific case magic goes into this). The next time such information is asked for, or other information which can be logically inferred from the previously collated information, it is sent directly from the cache, without consulting the remote servers. It provides multi-server group merging, and sophisticated weighted content-filtering, access control and a chrooted environment. It makes heavy use of shared memory, copy-on-write, fd caches, mmaping and more for efficiency capable of serving thousands of simultaneous users. To sum that up in layman's terms : nntpcache is a cache, similar to the Apache object-cache, but nntpcache works on newsgroups. By acting as a cache, nntpcache transparently and efficiently saves bandwidth enough to feed a small nation. It can also merge dozens of news servers into one large "virtual" news server, and provide content filtering (censoring! censoring!). It's a miserly beast with cpu and memory due to some gee-whizz trickery. -------------------------------------------------------------------------- Question 1.2. What operating systems does nntpcache support? nntpcache has been known to compile under these systems : * BSD FreeBSD 2.2-CURRENT + GCC2.6.3 BSDI [something] * LINUX Linux 1.3.77 + GCC 2.7.2 + libc5.2.18 * SOLARIS 2.4 + Native CC * HP-UX HP-UX 9.05 + HP ANSI C * OSF/1 If you find a OS/compiler/library tuple that nntpcache appears to compile and run under, then don't keep it to yourself, please let tell the devteam at nntpcache@nntpcache.org! (this list is.. err dated - nntpcache works on nearly all modern unixes) -------------------------------------------------------------------------- Question 1.3. Where can I find nntpcache? / Where is the latest version? The original archive is at : ftp://ftp.nntpcache.org/pub/nntpcache/nntpcache.tar.gz The UK, Europe and US mirror of nntpcache can be accessed at: ftp://planete.theplanet.net/pub/nntpcache/nntpcache.tar.gz The Norwegian and European mirror is at : ftp://ftp.nethelp.no/pub/news/nntpcache-1.0.7.1.tar.gz FreeBSD 2.2.x and above has a port/package available in the ports system in news/nntpcache Rumors on the wire suggest a Debian package also, but don't oink on it. -------------------------------------------------------------------------- Question 1.4. Can I use nntpcache freely or do I have to license? Probably. It depends on who you are. See the LICENSING file for full details, but in a nutshell, individuals, non-military, non-profit research groups, non-profit organisations (including educational institutions) and medical or educational government departments may use nntpcache freely. Everyone else (that is commercial organisations and government) are required to license in order to fund on-going development and support. -------------------------------------------------------------------------- Question 2.1. My system chokes on all that damned mmap()ing?!. What can I do? a) upgrade your OS. For instance, linux 1.2.13's mmap() does not work correctly in shared mode. Linux 2.0 rectified this. b) get a decent OS. Ultrix mmap is totally broken! -------------------------------------------------------------------------- Question 2.2. Netscape-news kills itself when reading news from nntpcache-0.88.3! We know. Didn't get a bugs-bounty either. Geesh. nntpcache-0.90+ formats its data in a way that makes netscape happier. But it IS netscape's fault for expecting nnrpd style response messages. -------------------------------------------------------------------------- Question 2.3. Expiring with nntpcached -e doesn't work Running nntpcached -e on a system that is already running nntpcache doesn't work. To expire on a system that is already running nntpcache, send the nntpcache process the USR2 signal. -------------------------------------------------------------------------- Question 2.4. Help! The size of our cache keeps growing! What should I do about expiration? Expiration should be automatic, it is controlled by the following in the config file: # expire if there are under this number of blocks in the cache # partition minBlocksFree 1000 # run expire if there are under this number of inodes in the cache # partition minFilesFree 1000 # don't let the history file grow larger than this (bytes) hisHighWater 60M # when it does, trim it back to this (bytes) hisLowWater 40M # When we do an expire, kill articles older than this maxArtAge 2w # check disk space, etc for automatic expiry this often # (but only after a client connects or disconnects) expireCheckPeriod 5m The important point that isn't made clear here is that expire runs only occur when minBlocksFree or minFilesFree are not satisfied. i.e the cache will happily grow until the disk starts filling up. Cache expiration isn't incredibly intelligent, although it is very fast and tries to track the actual state of affairs on the remote server. -------------------------------------------------------------------------- Question 2.5. nntpcache spits out : 'post failed From: address not in Internet syntax'. What does that mean? That error message translated into normal English means: Your attempt to post failed because the address in the "From:" field was not in the format user@host. To solve the problem, ensure you have a correct reply-to address configured in the newsreader. -------------------------------------------------------------------------- Question 2.6. Why does nntpcache-update take so long to load at initialisation? nntpcache update takes a while on master startup since it loads overview.fmt, active, newsgroups and active.times lists from the providing server(s). these lists typically run into several mega-bytes, and nntpcache isn't able to serve clients till it has its hot little hands on this vital information. -------------------------------------------------------------------------- Question 3.1. Is there a nntpcache mailinglist? Sure is! To join the thrills and spills of the mailinglist! Poke fun at Julian's coding style! Find bugs in Herbet Xu's bug fixes! Submit nntpcache filters to filter out all messages posted by the Kook of the Month and more! Send a message with the subject of 'subscribe' to the address 'nntpcache-users-request@nntpcache.org' (without the single quotes). -------------------------------------------------------------------------- Question 3.2. How do I unsubscribe from the mailinglist? Unsubscribing from the mailinglist is as simple as subscribing. To unsubscribe, send a message with the subject of 'unsubscribe' to the address 'nntpcache-users-request@nntpcache.org' (without the single quotes). -------------------------------------------------------------------------- Question 3.3. Is there an archive of the nntpcache mailinglist? Yes! There are actually two archives of the nntpcache mailinglist. The first archive is at : ftp://ftp.nntpcache.org/pub/nntpcache/mailinglist A HTML version of the mailinglist archive is available at: http://www.support.nl/online/nntpquery.html -------------------------------------------------------------------------- Question 3.4. Is there an official nntpcache home page? Is the pope catholic? The official nntpcache home-page is accessible at http://www.nntpcache.org/ -------------------------------------------------------------------------- Question 3.5. Who should I contact if I can't get nntpcache to work / I've found a bug in nntpcache, what should I do? If you're not subscribed to the mailinglist, send your report to 'nntpcache@nntpcache.org', otherwise send your report to the mailinglist. Along with your report, send your servers and access files and syslog output if appropiate to expedite a fix to your problem. ok. ok. expedite is a euphemism :) -------------------------------------------------------------------------- Question 4.1. What are all those entries in the servers configuration file? They signify how long nntpcache should keep the following -- > - Active the active file > - Act.tim active.times > - Newsgrp newsgroups > - Group What you get if you / a client types 'group some.newsgroup' > - Xover What you get if you / a client types 'xover low-high' > - Arts What you get if you / a client types 'head/body/article no' -------------------------------------------------------------------------- Question 4.2. How do I specify the "Xref:full" part in the "overviewFmt" line in the config file? You shouldn't have to specify the overviewFmt thing if the remote server supports 'LIST overview.fmt'. But if you insist, just 'Xref' will do. -------------------------------------------------------------------------- Question 4.3. I've entered a server but nntpcache doesn't seem to want to know about it. Have you made sure there's a period ('.') after the domain name of your server? For example: news.mynet.net DEFAULT 4h 2d 2d 30m 60d 60d is probably wrong, but: news.mynet.net. DEFAULT 4h 2d 2d 30m 60d 60d is likely correct. It isn't essential to use a period after the domain name ('.'), but it is essential that you use do so *consistently*. The period is used to anchor the domain to speed domain lookups, and nntpcache will treat references to a server without a period as an entirely different server to one with a period. If the server in questions is in /etc/hosts and you /etc/resolv.conf (or equivalent) tells your DNS resolver library routines to search /etc/hosts before NIS/DNS then you probably DON'T want the period appended (unless you have a corresponding period in the /etc/hosts entry) -------------------------------------------------------------------------- Q4.4 I want to set the nntpcache-virtual-server-size record, what are some public news-servers I can add to my servers file? >Does anyone have a list of corporate newsservers (ie: msnews.microsoft.com >for microsoft.*, news.3dfx.com for 3dfx.*, etc.)? If anyone can add to or has corrections for this list, please let me know. Some entries are commented because I haven't had time to specifically add all groups they carry or because they weren't responding. newsgroups.intel.com. DEFAULT 24h 4d 4d 30m 60d 60d msnews.microsoft.com. DEFAULT 24h 4d 4d 30m 60d 60d news.3dfx.com. DEFAULT 24h 4d 4d 30m 60d 60d miriam.fuller.edu. DEFAULT 24h 4d 4d 30m 60d 60d adesknews.autodesk.com. DEFAULT 24h 4d 4d 30m 60d 60d forums.borland.com. DEFAULT 24h 4d 4d 30m 60d 60d hope.harvard.net. DEFAULT 24h 4d 4d 30m 60d 60d cnews.corel.com. DEFAULT 24h 4d 4d 30m 60d 60d news.avault.com. DEFAULT 24h 4d 4d 30m 60d 60d news.linkexchange.com. DEFAULT 24h 4d 4d 30m 60d 60d #forums.macromedia.com. DEFAULT 24h 4d 4d 30m 60d 60d #news.activeie.com. DEFAULT 24h 4d 4d 30m 60d 60d #news.interaktiv-media.se. DEFAULT 24h 4d 4d 30m 60d 60d publicnews.msn.com. DEFAULT 24h 4d 4d 30m 60d 60d news.netobjects.com. DEFAULT 24h 4d 4d 30m 60d 60d secnews.netscape.com. DEFAULT 24h 4d 4d 30m 60d 60d beetle.privatei.com. DEFAULT 24h 4d 4d 30m 60d 60d massena.com. DEFAULT 24h 4d 4d 30m 60d 60d cor.newman.upenn.edu. DEFAULT 24h 4d 4d 30m 60d 60d forums.powersoft.com. DEFAULT 24h 4d 4d 30m 60d 60d news.sff.net. DEFAULT 24h 4d 4d 30m 60d 60d nntp.net-link.net. DEFAULT 24h 4d 4d 30m 60d 60d service.symantec.com. DEFAULT 24h 4d 4d 30m 60d 60d #jazz.trumpet.com.au. DEFAULT 24h 4d 4d 30m 60d 60d news.turbopower.com. DEFAULT 24h 4d 4d 30m 60d 60d news.ksc.nasa.gov. DEFAULT 24h 4d 4d 30m 60d 60d news.glbs.com. DEFAULT 24h 4d 4d 30m 60d 60d %BeginGroups # Group pattern Host * your.main.server.here alt.oui.support* beetle.privatei.com. asp.* news.glbs.com. audioweb.* news.glbs.com. 3dfx.* news.3dfx.com. autodesk.* adesknews.autodesk.com. avault.* news.avault.com. borland.* forums.borland.com. cakewalk.* hope.harvard.net. corel.* cnews.corel.com. crescent.* news.glbs.com. db.* news.glbs.com. forum.* news.avault.com. glbs.* news.glbs.com. intel.* newsgroups.intel.com. le.* news.linkexchange.com. linux.* miriam.fuller.edu. microsoft.* msnews.microsoft.com. msn.* publicnews.msn.com. net-link.* nntp.net-link.net. netobjects.* news.netobjects.com. netscape.* secnews.netscape.com. ntworthy* news.glbs.com. people.* secnews.netscape.com. pilot.* massena.com. powerbasic.* news.glbs.com. powersoft.* forums.powersoft.com. sff.* news.sff.net. sybase.* forums.powersoft.com. symantec.* service.symantec.com. #trumpet.* jazz.trumpet.com.au. turbopower.* news.turbopower.com. woll2woll* news.glbs.com. mail.winvn news.ksc.nasa.gov. perl.porters-gw cor.newman.upenn.edu. -------------------------------------------------------------------------- Acknowledgements: Alan Brown <alan@manawatu.gen.nz> Herbert Xu <herbert@greathan.apana.org.au> Julian Assange <proff@nntpcache.org> Ken Eves <cyberw4@cyberworldcafe.com> --- We've heard that a million monkeys at a million keyboards could produce the Complete Works of Shakespeare; now, thanks to the Internet, we know this is not true. --Robert Wilensky, University of California