coolidge@brutus.cs.uiuc.edu (John Coolidge) (10/12/89)
chuq@Apple.COM (Chuq Von Rospach) writes: >>brodie@moocow.mcw.edu requested the sys file of every site in the >>world with a "sendsys" control message. >Gads. I hope he has a *large* disk in /usr/spool/mail. Actually, he's on a VMS machine, so the mail goes directly into his mail folder analogue. He _might_ be safe if on a quota'd disk, since overquota mail might just be bounced --- unless mail has exquota privs. In that case, it's hello full user disk, and probably bye bye any mail (since the only reasonable recovery is mass delete). This brings up a point: if you just want to check connectivity, the 'version' control is almost certainly a better option. Or 'senduuname'. 'sendsys' should be reserved for cases when you _really_ want a sys file (for debugging a feed, for instance), and should be carefully checked for the right Distributions: and Newsgroup: headers before releasing it into the world. 'version' and 'senduuname' both give you just one line of output (plus the obligatory massive headers). 'sendsys' gives you lots of lines for each site... --John -------------------------------------------------------------------------- John L. Coolidge Internet:coolidge@cs.uiuc.edu UUCP:uiucdcs!coolidge Of course I don't speak for the U of I (or anyone else except myself) Copyright 1989 John L. Coolidge. Copying allowed if (and only if) attributed. You may redistribute this article if and only if your recipients may as well.
mcb@ncis.tis.llnl.gov (Michael C. Berch) (10/13/89)
In <1989Oct11.235318.23465@brutus.cs.uiuc.edu> coolidge@cs.uiuc.edu writes: > chuq@Apple.COM (Chuq Von Rospach) writes: > >>brodie@moocow.mcw.edu requested the sys file of every site in the > >>world with a "sendsys" control message. > > >Gads. I hope he has a *large* disk in /usr/spool/mail. > > Actually, he's on a VMS machine, so the mail goes directly into his mail > folder analogue. He _might_ be safe if on a quota'd disk, since overquota > mail might just be bounced --- unless mail has exquota privs. In that > case, it's hello full user disk, and probably bye bye any mail (since > the only reasonable recovery is mass delete). Huh? Am I missing something? According to the newsstats posting from uunet last week, there were about 5,100 sites posting news. The arbitron stats estimate a total of about 12,000 sites. That's an upper bound. I'm willing to wager that no more than 7,500 sites are able or willing to respond to a sendsys message. I have no clue as to the average size of a sys file, but ours is 2K bytes and is horrendously verbose, with contact info for every feed we've had in the last couple of years. I constructed a couple of "normal" sys files, and they seemed to be about 800 bytes -- let's round up to 1K. Thus, 7500 * 1024 = 7,680,000 ~= 7.7 MB. I'd say for all but the home systems or PCs, an extra 7.7MB of mail over a few days ain't going to fill up too many disks. At most places I'm familiar with it probably wouldn't be *noticed*, disk space-wise, though I'm sure the size of the mailer log might attract attention. -- Michael C. Berch mcb@tis.llnl.gov / uunet!tis.llnl.gov!mcb
coolidge@brutus.cs.uiuc.edu (John Coolidge) (10/13/89)
mcb@ncis.tis.llnl.gov (Michael C. Berch) writes: >In <1989Oct11.235318.23465@brutus.cs.uiuc.edu> coolidge@cs.uiuc.edu writes: >> chuq@Apple.COM (Chuq Von Rospach) writes: >> >Gads. I hope he has a *large* disk in /usr/spool/mail. >> >> Actually, he's on a VMS machine, so the mail goes directly into his mail >> folder analogue. He _might_ be safe if on a quota'd disk, since overquota >> mail might just be bounced --- unless mail has exquota privs. In that >> case, it's hello full user disk, and probably bye bye any mail (since >> the only reasonable recovery is mass delete). >Huh? Am I missing something? According to the newsstats posting >from uunet last week, there were about 5,100 sites posting news. >The arbitron stats estimate a total of about 12,000 sites. That's an >upper bound. I'm willing to wager that no more than 7,500 sites are >able or willing to respond to a sendsys message. I have no clue as to >the average size of a sys file, but ours is 2K bytes and is >horrendously verbose, with contact info for every feed we've had in >the last couple of years. I constructed a couple of "normal" sys >files, and they seemed to be about 800 bytes -- let's round up to 1K. >Thus, 7500 * 1024 = 7,680,000 ~= 7.7 MB. I'd say for all but the >home systems or PCs, an extra 7.7MB of mail over a few days ain't >going to fill up too many disks. At most places I'm familiar with it >probably wouldn't be *noticed*, disk space-wise, though I'm sure the >size of the mailer log might attract attention. An extra 7.7MB of mail would blow our mail partition totally out of the water (of course, that's because it's way too small, but that's beside the point; it does happen). I have a feeling that most of the 9000ish sites in the comp.mail.maps list (vaguely remembered number, which is I think fairly accurate) will respond to a sendsys. I'd guess the number of responses to be at least 8000. Of course, what would be more valuable would some stats on what this sendsys actually produced... (of course, that confuses the issue with facts :-) ). But more importantly, 1024 bytes is _way_ too small a response pattern. Our sys file is around 4K. I've got the sys files of several of my connections on file, and the smallest was 3k, with the largest at 14K. Granted that I talk to large sites (I've got 12 connections, and my largest connection has 40), there are still a fair number of sites with five-ten connections around. 1K won't handle that many, not if anywhere near a broad range of groups are carried, or if any special batch files, etc., are specified. And all this is still ignoring all of the mail headers. For an average site, internet-internet transmission will entail around 400-800 bytes of mail headers. Once you go off the internet and into uucp, these group rapidly (more Received: headers, which are the big space user). I've seen 10K or larger mail headers, although these are rare. Considering that moocow is one hop off the internet (I think), and that most of the >8000 responders will be off the internet and not within short uucp range of moocow, mail headers alone will probably average in the 1K range. I'd guess that 2.5K*8000, or 20M, is a closer ballpark figure. It could still be short. 20M would put a serious crimp on many people's mail spool areas, I think. It would certainly mess us up if we were just carrying it for transmission, and it would totally overflow our mail disk if someone here was getting it. Again, I'd love some real numbers from moocow when all this settles down (average file size, number of sys files, etc.) And all of this ignores all of the letters of protest that could arise when sites find they're moving multiple megabytes of sys files around all headed for one site. (And, of course, the added news volume from discussing the added volume :-) ). And the poor link to moocow! This sort of volume will probably double the load on that link for a few days. --John -------------------------------------------------------------------------- John L. Coolidge Internet:coolidge@cs.uiuc.edu UUCP:uiucdcs!coolidge Of course I don't speak for the U of I (or anyone else except myself) Copyright 1989 John L. Coolidge. Copying allowed if (and only if) attributed. You may redistribute this article if and only if your recipients may as well.
asp@uunet.UU.NET (Andrew Partan) (10/13/89)
In article <1989Oct13.125402.17356@brutus.cs.uiuc.edu>, coolidge@brutus.cs.uiuc.edu (John Coolidge) writes: > Our sys file is around 4K. I've got the sys files of several of my > connections on file, and the smallest was 3k, with the largest at 14K. uunet's sys file: 176 -rw-rw-r-- 1 news 166812 Oct 13 09:40 /usr/spool/news-lib/sys This is probably the largest -:). We are feeding 404 sites & counting. --asp@uunet.uu.net (Andrew Partan) ASN.1 Object Identifier: "{joint-iso-ccitt mhs(6) group(6) 157}"
rsalz@bbn.com (Rich Salz) (10/14/89)
Back before June Usenix I put out a sendsys with Inet distribution and cancelled it 24 hours later. I got responses from 297 sites, for a total of 720K (including mail headers). Discounting big sites, the average response was about 2500 bytes. The big sites I discounted were: 10821 bloom-beacon.mit.edu 11016 news.arc.nasa.gov 12067 prc.unisys.com 19596 rutgers.edu 15919 tut.cis.ohio-state.edu 20355 ucbvax.berkeley.edu 12605 ms.uky.edu 142456 uunet.uu.net I think your average INET site is "bigger" than your average non-Inet site. /rich $alz -- Please send comp.sources.unix-related mail to rsalz@uunet.uu.net. Use a domain-based address or give alternate paths, or you may lose out.
coolidge@brutus.cs.uiuc.edu (John Coolidge) (10/14/89)
asp@uunet.UU.NET (Andrew Partan) writes: >In article <1989Oct13.125402.17356@brutus.cs.uiuc.edu>, coolidge@brutus.cs.uiuc.edu (John Coolidge) writes: >> Our sys file is around 4K. I've got the sys files of several of my >> connections on file, and the smallest was 3k, with the largest at 14K. >uunet's sys file: > 176 -rw-rw-r-- 1 news 166812 Oct 13 09:40 /usr/spool/news-lib/sys I said those were the largest I had around, not the largest in the world! :-) I knew that uunet's sys file was about as long as they come. It'll be a while before anyone else comes close, I think... --John -------------------------------------------------------------------------- John L. Coolidge Internet:coolidge@cs.uiuc.edu UUCP:uiucdcs!coolidge Of course I don't speak for the U of I (or anyone else except myself) Copyright 1989 John L. Coolidge. Copying allowed if (and only if) attributed. You may redistribute this article if and only if your recipients may as well.
davidc@vlsisj.VLSI.COM (David Chapman) (10/14/89)
In article <1989Oct11.235318.23465@brutus.cs.uiuc.edu> coolidge@cs.uiuc.edu writes: <chuq@Apple.COM (Chuq Von Rospach) writes: <>>brodie@moocow.mcw.edu requested the sys file of every site in the <>>world with a "sendsys" control message. < <>Gads. I hope he has a *large* disk in /usr/spool/mail. < <Actually, he's on a VMS machine, so the mail goes directly into his mail <folder analogue. He _might_ be safe if on a quota'd disk, since overquota <mail might just be bounced --- unless mail has exquota privs. In that <case, it's hello full user disk, and probably bye bye any mail (since <the only reasonable recovery is mass delete). I don't know about net mail bouncing, but at our site under DECnet (local) the message simply isn't sent when the recipient's quota is exceeded, so we don't have to worry about one user taking over the whole disk. I don't know if you can change mail's privileges under VMS, so I can't guarantee his safety on his system. :-) -- David Chapman {known world}!decwrl!vlsisj!fndry!davidc vlsisj!fndry!davidc@decwrl.dec.com
mrm@sceard.Sceard.COM (M.R.Murphy) (10/15/89)
In article <70024@uunet.UU.NET> asp@uunet.UU.NET (Andrew Partan) writes: >In article <1989Oct13.125402.17356@brutus.cs.uiuc.edu>, coolidge@brutus.cs.uiuc.edu (John Coolidge) writes: >> Our sys file is around 4K. I've got the sys files of several of my >> connections on file, and the smallest was 3k, with the largest at 14K. > >uunet's sys file: > 176 -rw-rw-r-- 1 news 166812 Oct 13 09:40 /usr/spool/news-lib/sys > >This is probably the largest -:). We are feeding 404 sites & counting. > Speaking of the sendsys control message, and the version control message that came a bit afterwards, the response is mailed to the news path in the article by the version of news running here. Since, as we all know, the connectivity of the news path is not necessarily e-mail connectivity, a bounce is likely. I reckon I'll either set up our mailer for ERR (Extremely Rabid Rerouting), or go and fix news to respond somewhere other than the news path for sendsys and version. Maybe to the Reply-To: address? Maybe I should already have done this? :-) How about some comment from the sites just upstream of moocow on increase in load? Same for major sites fed by uunet? -- Mike Murphy Sceard Systems, Inc. 544 South Pacific St. San Marcos, CA 92069 mrm@Sceard.COM {hp-sdd,nosc,ucsd,uunet}!sceard!mrm +1 619 471 0655
mcb@ncis.tis.llnl.gov (Michael C. Berch) (10/17/89)
[Re: what happens when you send out a global sendsys...] Thanks to John Coolidge for pointing out what I *did* miss in the original calculation: the size of the (expanding) mail header. That indeed could well add up to 1K by itself for a site a couple of hops off the Internet. I'd also accept the 8,000 sites vs. 7,500 sites estimate, though I think we're both speculating. But I really think that 1K is probably a good average sys file, since for every site that has 5-10 feeds there are several that have only 1 or 2, which might result in a 400-500 byte sys file (a hundred-char line for each feed, plus a line of contact info per site, plus a couple of lines of comments). So then we get 8,000 * 2048 = 16,384,000 ~= 16.4 MB You would hope that wouldn't blow too many mail spool disks, but perhaps so. The real squeeze, the one I forgot since we do not run UUCP on our news host, is the space required for pass-through mail. -- Michael C. Berch mcb@tis.llnl.gov / uunet!tis.llnl.gov!mcb
clarke@acheron.uucp (Ed Clarke/10240000) (10/17/89)
From article <547@ncis.tis.llnl.gov>, by mcb@ncis.tis.llnl.gov (Michael C. Berch): > [Re: what happens when you send out a global sendsys...] > 8,000 * 2048 = 16,384,000 ~= 16.4 MB > You would hope that wouldn't blow too many mail spool disks, but perhaps > so. The real squeeze, the one I forgot since we do not run UUCP Last year (December) I sent out a 'sendsys' with a NY distribution. Somehow it multiplied so that rochester.edu got 22 requests from me within a span of ten hours ... I still haven't figured out how it happened and have no intention of trying it again. So you only hope it's 16 meg - -- Ed Clarke acheron!clarke
richard@gryphon.COM (Richard Sexton) (10/18/89)
In article <1989Oct17.045700.11226@acheron.uucp> clarke@acheron.uucp (Ed Clarke/10240000) writes: >From article <547@ncis.tis.llnl.gov>, by mcb@ncis.tis.llnl.gov (Michael C. Berch): >> [Re: what happens when you send out a global sendsys...] >> 8,000 * 2048 = 16,384,000 ~= 16.4 MB >> You would hope that wouldn't blow too many mail spool disks, but perhaps >> so. The real squeeze, the one I forgot since we do not run UUCP > >Last year (December) I sent out a 'sendsys' with a NY distribution. >Somehow it multiplied so that rochester.edu got 22 requests from me >within a span of ten hours ... I still haven't figured out how it >happened and have no intention of trying it again. Perhaps there ought to be a limit in the software viz. how many times per day (hour?) it will respond to a sendsys. 100 sendsys requests per hour could get pretty annoying. -- Help wipe out BBQ lighter fluid in your lifetime richard@gryphon.COM decwrl!gryphon!richard gryphon!richard@elroy.jpl.NASA.GOV