Article View: chi.mail
Article #849CLOUT Tips and Notes
From: Dave Ihnat
Date: Thu, 21 May 1992 21:00
Date: Thu, 21 May 1992 21:00
129 lines
7550 bytes
7550 bytes
Hi, folks... I'm one of the people involved with the /usr/group/chicago CLOUT project; more specificially, I own/operate one of the two CLOUT nodes (homebru/clout), and have been involved in the project since its inception. In the course of things, I occasionally catch some of the questions/complaints, and am at the bottom of some of the events--good *and* bad--that affect delivery, as well as participating in the organization meetings. As I come up with items that are of general interest, I'd like to "share them" with you (Note new corporate buzz word.) Please note that there is an official organ of /usr/group/chicago which issued similar updates and information here; I'm the unofficial one. Any time there is a difference of opinion expressed between my postings and those of /usr/group, please accept theirs as gospel (and expect there'll be a fine donnybrook at the next meeting...) 1) NEW HOMEBRU/CLOUT ==================== This is the top item that drove this posting. I just (in the last few weeks) fully rebuilt homebru/clout from a 3b1 running its own Unix 3.51m on a 140Mb MFM drive, to a 386DX/33 with a 340Mb SCSI running Dell SVR4. This has not, unfortunately, been without some outages and mailer problems; for this, I apologize. The most recent gotcha occurred yesterday and part of today. (NOTE: For anyone installing smail 3.1 under Dell Unix in particular, and HDB UUCP in general--don't forget to make sure that the Grades file has an entry for the default grade issued by smail. It doesn't default to a low grade or anything so nice; it rejects the job.) At this point, I believe the system is stable; please give it two or three days for any problem messages to clear, then please let me know if you experience any problems that seem to originate on homebru/clout. As usual, postmaster@clout.chi.il.us is the first-line for reporting problems, so simply CC to me as ignatz@homebru.chi.il.us if you feel it's in my baliwick. 2) NET MAPS =========== Recently, the administrator of a well-connected and heavily used CLOUT client received a complaint from one of the other CLOUT clients, to the effect that they figured that his system was redirecting mail to their site. Why? Their mail was being routed to a site in *California*! Examination of the problem proved this wasn't the issue. Rather, there's a bit of confusion out there concerning what should or should not be submitted to the UUCP Mapping Project for CLOUT clients. The rules are actually fairly simple: 1) Normal CLOUT clients do NOT submit a map entry. They are described in the CLOUT map, and only there. Subsequently, all mail directed to any machine in the chi.il.us domain will end up at CLOUT, which knows how to forward to that machine. 2) IF you need to submit a map--for instance, you have connections other than just your link to CLOUT, and you wish to advertise them to the world--you should only list a fully qualified domain name on the line where you declare your connections, i.e.: (a) homebru.chi.il.us vfrot(DAILY), chinet(DEMAND), etc. 3) DO NOT EQUATE YOUR MACHINE NAME WITH THE FULLY QUALIFIED NAME UNLESS YOU'VE RECEIVED ASSURANCE FROM THE MAPPING PROJECT THAT YOUR NAME IS UNIQUE! In fact, don't even trust this--scan the entire map database to make sure the name you've chosen isn't declared by some other site!! An example of this follows (the machine name has been changed to protect me from being yelled at by the real machine owner): (b) foobar = foobar.chi.il.us foobar homebru(HOURLY), etc. In the problem case mentioned above, it turned out that some site in California advertised a link to a machine named 'foobar', that had no map entry of its own. Since 'foobar' was equated to 'foobar.chi.il.us', as in example (b), despite the fact they had a unique address in the message header, they suddenly were equated with the machine half a continent away. This wouldn't have happened if they gave the fully qualified name as in example (a). The CLOUT Administration Group will, in the future, perform periodic audits of their clients, at least, to catch problems like this; but this advice applies to ANYONE who submits a map. 3) NETWORK ROUTING ================== In another case, the administrator of a CLOUT client machine complained somewhat bemusedly that despite the fact that he *knew* that the CLOUT node to which he was attached, at the University of Chicago, was directly on the Internet and could drop mail right at the (virtual) door of 'uunet', any mail originating on his system for uunet took a circuitous route through 'mcdchg' and about 4 other machines. Why? The problem in this case was the fact that he was running a full map database on his machine. The CLOUT machine at the University of Chicago runs smail 3.1, and 'knows' that it has several means of delivery, including SMTP over the Internet. His machine knows nothing that isn't in the UUCP maps, and the costs associated with those maps don't reflect direct Internet links. The solution in his case--and in the case of any CLOUT client in a similar situation--is to replace the full paths database with a truncated database reflecting his first-hop UUCP neighbors, and to declare their CLOUT machine to be a 'smart host'. This is supported in smail 2.5, smail 3.1, ELM, and various other MTA/MUA combinations. Thus, any address to which your machine has no direct link will be passed to the CLOUT node, which will attempt to deliver first over the Internet, then as indicated by its copy of the PATHS database. This has two advantages for the CLOUT client. First, large chunks of convoluted address paths can be eliminated, as all machines on the Internet are effectively one hop away from each other. Secondly, the not-inconsiderable overhead--in both disk space, CPU cycles, and administrator time--of maintaining a private copy of the paths database is handed off to the CLOUT project. An alternative scheme, should you need to use a local copy of the full database, is to use smail 3.1 and create a second UUCP entry in your routers file. Preceed the general UUCP paths file with a custom local copy that advertises known Internet hosts as **very** cheaply connected, and you link to the CLOUT host as even cheaper. In this manner you can force direction through the Internet. Hope this has been a useful bit of information...if I hear that nobody liked it, I'll stop wasting your time. Otherwise, I'll do it again as I come up with interesting happenings. Dave Ihnat ignatz@homebru.chi.il.us (preferred return address) ignatz@chinet.chi.il.us
Technical Details
Message-ID:
<1992May21.210007.6241@chinet.chi.il.us>Path:
arc1-ca.pugleaf.net!sparky!uunet!tellab5!nucsrl!news.acns.nwu.edu!news.ils.nwu.edu!uxa.ecn.bgu.edu!uchinews!machine!chinet!ignatzShow Full Headers
Complete Article Headers
X-Google-Language: ENGLISH,ASCII-7-bit X-Google-Thread: 10e835,16c2b4207adc706,start X-Google-Attributes: gid10e835,public X-Google-Thread: f8d04,16c2b4207adc706,start X-Google-Attributes: gidf8d04,public X-Google-ArrivalTime: 1992-05-21 14:00:07 PST Newsgroups: chi.mail,chi.general Path: arc1-ca.pugleaf.net!sparky!uunet!tellab5!nucsrl!news.acns.nwu.edu!news.ils.nwu.edu!uxa.ecn.bgu.edu!uchinews!machine!chinet!ignatz From: ignatz@chinet.chi.il.us (Dave Ihnat) Subject: CLOUT Tips and Notes Summary: Some bits of information and news on /usr/group/chicago's CLOUT Message-ID: <1992May21.210007.6241@chinet.chi.il.us> Date: Thu, 21 May 1992 21:00:07 GMT Distribution: chi Organization: Chinet - Public Access UNIX Keywords: /usr/group/chicago Email CLOUT Lines: 129