Path: utzoo!utgpu!jarvis.csri.toronto.edu!mailrus!cs.utexas.edu!usc!zaphod.mps.ohio-state.edu!rpi!tale From: tale@cs.rpi.edu (David C Lawrence) Newsgroups: news.newusers.questions Subject: Re: How can I access USENET articles database on my computer ? Message-ID: Date: 10 Feb 90 06:27:38 GMT References: <8000@cbnewsh.ATT.COM> <8006@cbnewsh.ATT.COM> Organization: Rensselaer Polytechnic Institute, Troy NY Lines: 101 In <8006@cbnewsh.ATT.COM> wcs@cbnewsh.ATT.COM (Bill Stewart) writes: > There are two main ways netnews gets passed around: > - The B News / C News method > - The NNTP Network News Transfer Protocol. This is an inaccurate way to characterise the system. NNTP is meant to be exclusive of the resident news system. The above makes it sound as though you can't use NNTP with either of the above systems, which is wholly fallacious. The rest of this article will address things in the Unix world, since as you mention these are the main news machines. There are many other methods of storage/retrieval/transport in use. > In B/C News, all news articles that a machine might want get shipped to > that machine, using whatever network is around (uucp, ftp, etc.). If anyone at all is using FTP to regularly pass news as part of a feed, I'd really like to hear more about it. Neither uucp nor ftp are strictly networks, by the way. > (Article numbers are different on each machine. Well, mostly true. Depends on how you want to semantically look at this. I would say "at each site" and call the amalgomation of all machines which share a single news database one site. > The command "hgrep" is a grep that only greps article headers and > skips the body - much faster.) It is also a non-standard grep. While easy enough to write, it is not available on many, many machines. Certainly a much smaller number than have grep. > NNTP is a different approach, designed for use in a TCP/IP network. It is not a different approach to storage and feed arrangement. It is only a different way of passing articles over the line. When RPI had a couple of different machines hosting news, nearly all used NNTP to get thier news. One also ran a small UUCP feed. All of them used B News. Since we centralised I also moved us to C News. rpi.edu feeds and receives all of its news, even locally posted articles, via NNTP. > Essentially, you'll have a server machine that feeds a bunch of others, > which distributes a certain amount of the database of what articles it has. One very important thing to remember is that NNTP does double duty as both a transport mechanism between feeds and as a database retrieval system for news readers. > When a client machine wants an article (because some user wants to > read it), it retrieves the article from the server. I don't know if > the client keeps the article around for a while or not. It depends very much on the client. If the client is another news site, it keeps the article around until it expires it. If it is a news reader then it is up to the reader how it wants to interact with the server and its user. Most common NNTP readers now only keep the full text of one article at any given time. > NNTP is much more efficient, and is the way much of the Internet gets > its news, but requires teaching the newsreaders how to use it. Not necessarily, but you've covered that base in the next paragraph. > An intermediate approach is to mount /usr/spool/news over a remote > file system like RFS or NFS, with a few hacks to inews to make > outgoing news do the right thing. This is less efficient than NNTP > (being an NFS server is more work than an NNTP server), but is > pretty transparent. rpi.edu uses both methods. rpi.edu:/usenet is exported read-only via NFS to anyone on campus that wants it and can take it. All of the USENET operation for that machine as a server is kept under /usenet, including the source code, spool files, client support for various architectures, and administrative tools. Readers can use either NFS or NNTP. In <859@dftsrv.gsfc.nasa.gov> packer@chrpserv.gsfc.nasa.gov (Charles Packer): > The disadvantage of the NNTP method is that the message database is > not available for perusal by means other than the news reader, > since the files reside on a different computer than the user's. If the reader is up to the challenge then it should be better than anything except truly local spooling for most scanning (as grep(1) would do) and paging (as more(1) would do). I basically agree with you though; if I have the spool mounted on my machine I would (and do) occasionally go right to it in a shell rather than bothering to fire up my news reader. > At least one progammer on the net is developing a modified version > of the nntp reader that simply sucks messages off the server into > your computer. Does he know about the existing nntpxfer which comes as part of the NNTP distribution? While admitted by its author to be a hack, it does exactly that. -- (setq mail '("tale@cs.rpi.edu" "tale@ai.mit.edu" "tale@rpitsmts.bitnet")) "Nice plant. Looks like a table cloth."