ptownson@bu-cs.BU.EDU (Patrick Townson) (12/07/88)
Seeing the recent messages regards the steadily increasing volume of traffic on the net leads me to believe that before long net etiquette may need to be rewritten to keep us all from getting buried in news. [I say what follows because I enjoy being such a popular fellow :) ] The machinery obviously does not know the difference between *news* and *signatures*. Nor does it know the difference between new news and a quote from the article being commented on. It has also been observed by astute net-watchers that I don't know the difference between my ass and my elbow, but that is something we can discuss at a later time. May I respectfully suggest that cutting signatures to their bare bones and *greatly* limiting the use of quotes from previous messages might reduce traffic and associated storage requirements by one third? I realize the problems which arise from posting without some reference to the earlier message. Without some point of reference, it can be impossible for the reader to know what you are replying to, or to whom. But still, it seems ridiculous to include entire messages in new messages. Some of the quotes on the net remind me of the situation with Jean Paul Sarte. When he was asked to write an introduction to a very short book by Jean Genet, he wrote a 500 page treatise to sit on the front end of a book about 150 pages long! My rule of thumb -- if I must quote at all -- is to write one or two introductory sentences of my own describing what I am responding about. They usually take the form -- "So.and.so from site.place wrote saying blah blah, and da da. He noted this and also that. I agree, because (rest of response here)... (or) he is wrong, because (rest here)....." I realize some messages require somewhat more elaborate inclusion of prior remarks, but frankly, most do not. We are not writing legal papers which have to be filed in quintuplicate with some government agency. I would never include message reference numbers in quotes simply because they are meaningless to anyone not on that machine. Inserting someone else's fully qualified network address is rarely needed. 'joe.blow@schmoe' is usually sufficient to refresh the memory of others who read your post. Yet these lines with reference numbers and full address routings take up (apparently) badly needed space on many machines. Why not try writing creative replies incorporating as little as required of other people's text right in your own message? As for signatures, yes they look lovely, but they may be getting too expensive from a machine resource point of view. I suggest still another group on the net, which for lack of a better name I will call net.contributors. This file can hold signatures, 'thumb-nail' biographical sketches of anyone who wants to be listed, net addresses; postal addresses; phone numbers; corporate affiliations; whatever. Anyone who thinks you (or I) are that terribly, terribly witty and intelligent that they want to make contact can review the net.contributors index. For the posting itself, just sign the thing, like I always do, and will do now. Patrick Townson PS: If a postscript is necessary or desirable, keep it short and succinct, like this one. As for those damnable disclaimers, perhaps a future revision to readnews could issue a blanket disclaimer message each time someone 'tuned in' on their machine. Something to the effect, 'the opinions expressed herein for the remainder of your session in readnews are solely that of the author of each item and are not to be construed as the opinion or position of the proprietor of the originating machine, or the network itself.' Then we could eliminate the thousand or so lines of disclaimers which have to pass over the wire each day. Thanks for thinking about it, anyway! Patrick
bill@ssbn.WLK.COM (Bill Kennedy) (12/08/88)
In article <26469@bu-cs.BU.EDU> ptownson@bu-cs.BU.EDU (Patrick Townson) writes: [ he said cut it down, so I did ] > >Thanks for thinking about it, anyway! > >Patrick There's another thing that would help a lot. If the leaves only take groups that people actually read then that amount of traffic is eliminated. The upstream sites could coordinate with the sites they feed and only carry the groups actually read downstream and locally. My newsfeed has awarded "the most obfuscated sys line" prize but that's what we did here. If someone here reads the group or someone downstream wants it then we carry it. There seems to be little reason or justification for all sites to carry a full feed but a lot do. It takes some coordination and cooperation among the admins but it results in better utilization of the resources. Obviously some site in each major area has to have a full feed in order to distribute what their neighbors want. If we started at the leaves (ssbn is really a leaf even though we feed a couple of sites) and made up the "obfuscated sys line" and sent it up, then our feeds could consolidate them and cut out the groups they don't need to carry. That could float all the way up to the "newsbones", recovering bandwidth and disk space all the way up. Elimination of articles that should be mailed is not necessarily a good solution. If the text is of interest to more than just the recipient then it's more efficient to post. Until we can apply the techniques to mail that we use for news (compression, batching, etc.) then posting for more than one addressee still appears to be a win. Speaking for only this site, I would be glad to spend the cycles to uncompress/unbatch mail and to recompress/rebatch pass-thru traffic in order to gain the improved utilization of the phone lines. If you're long distance from 99.9% of your neighbors, ssbn is, then it doesn't make a lot of sense to run a uucico for a single mail message unless it's the only thing in the queue. Patrick's suggestions are good, and they will certainly help if they are followed. I just wanted to point out that there are site-wide opportunities to improve utilization as well. -- Bill Kennedy usenet {killer,att,cs.utexas.edu,sun!daver}!ssbn!bill internet bill@ssbn.WLK.COM
karl@triceratops.cis.ohio-state.edu (Karl Kleinpaste) (12/08/88)
bill@ssbn.WLK.COM (Bill Kennedy) writes:
There's another thing that would help a lot. If the leaves only take groups
that people actually read then that amount of traffic is eliminated...
My newsfeed has awarded "the most obfuscated sys line" prize...
...If we started at the leaves (ssbn is
really a leaf even though we feed a couple of sites) and made up the
"obfuscated sys line" and sent it up, then our feeds could consolidate them
and cut out the groups they don't need to carry.
While I think the idea is very good, in practice it causes other
problems of its own. Notable, from my perspective, is the trouble
caused by rnews spending lots of time parsing the sys file. Our
limited resource is mostly CPU (and phone lines; 10 isn't enough for
the number of news neighbors and anon uucp'ers we've got), and I get
really nervous when I see osu-cis unbatching simultaneously from all
of att, killer, and tut.cis.ohio-state.edu, all while news is being
rebatched with compression to the leaf sites, because the CPU is being
flogged into the ground - load, as measured from a utility similar to
BSD's `uptime,' shows numbers in the vicinity of 11-14 in such cases.
Rnews takes over the system.
We actually had to request that one particular site, which had an
extremely convoluted sys line on osu-cis, to move on and take a feed
elsewhere around town; his sys line was huge, but contained exactly
the set of groups he wanted. The tradeoff was not acceptable. The
removal of that one site (18 others remain) was enough to leave the
system responding acceptably. Reducing the amount of news sent didn't
increase the ability of my system to feed news.
It's more efficient for massfeed sites to have all downstream feeds be
relatively simple in terms of distribution, as opposed to worrying
about the Mbyte volume going over the lines.
--Karllmb@vsi1.UUCP (Larry Blair) (12/08/88)
In article <KARL.88Dec7153323@triceratops.cis.ohio-state.edu> karl@triceratops.cis.ohio-state.edu (Karl Kleinpaste) writes: =bill@ssbn.WLK.COM (Bill Kennedy) writes: = There's another thing that would help a lot. If the leaves only take groups = that people actually read then that amount of traffic is eliminated... = =While I think the idea is very good, in practice it causes other =problems of its own. Notable, from my perspective, is the trouble =caused by rnews spending lots of time parsing the sys file. Our =limited resource is mostly CPU (and phone lines; 10 isn't enough... While I'm not sure whether parsing long sys file lines takes much cpu, I do know that batching and compressing for downstream leaves does. For this reason we run a group batch (thanks to csg@pyramid) which allows us to batch once for most of our leaves. It saves a lot of cpu and spool space. It doesn't lend itself to selective batching, though, and it means you'll have to add another sys file line if you want ihave/sendme. -- Larry Blair ames!vsi1!lmb lmb%vsi1.uucp@ames.arc.nasa.gov
jfh@rpp386.Dallas.TX.US (The Beach Bum) (12/08/88)
In article <KARL.88Dec7153323@triceratops.cis.ohio-state.edu> karl@triceratops.cis.ohio-state.edu (Karl Kleinpaste) writes: >It's more efficient for massfeed sites to have all downstream feeds be >relatively simple in terms of distribution, as opposed to worrying >about the Mbyte volume going over the lines. One last item - I have no qualms about telling a downstream site to feed someone else. I have yet to run into a situation where someone refused. Twigs and leafs may be better able to handle Bill's obfuscated sys line by virtue of seeing a smaller slice of the entire news stream. Perhaps that is what is saving me - I don't take talk or soc ;-) -- John F. Haugh II +-Cat of the Week:--------------_ /|- VoiceNet: (214) 250-3311 Data: -6272 |Aren't you absolutely sick and \'o.O' InterNet: jfh@rpp386.Dallas.TX.US |tired of looking at these damn =(___)= UucpNet : <backbone>!killer!rpp386!jfh +things in everybody's .sig?-------U---
henry@utzoo.uucp (Henry Spencer) (12/09/88)
In article <26469@bu-cs.BU.EDU> ptownson@bu-cs.BU.EDU (Patrick Townson) writes: >As for signatures, yes they look lovely, but they may be getting too >expensive from a machine resource point of view.... >... For the posting itself, just sign the >thing, like I always do, and will do now. > >Patrick Townson Unfortunately, signatures containing network address, not just name, are a practical necessity if you want people to be able to get mail to you. The article-header information is *not* sufficient except in very favorable cases. The policy I follow is that it takes two lines to give my name, affiliation, and network addresses. Including physical address and phone number is pointless, so there's no need to clutter up the signature with them. (Actually, even affiliation is kind of marginal.) Anything else I can cram into those two lines is fair game. :-) :-) At least my signatures get reactions -- I don't think there's been one in the last six months that hasn't at least drawn mail, and some of them have started real uproars among the narrowminded. -- SunOSish, adj: requiring | Henry Spencer at U of Toronto Zoology 32-bit bug numbers. | uunet!attcan!utzoo!henry henry@zoo.toronto.edu
pjh@mccc.UUCP (Pete Holsberg) (12/09/88)
In article <KARL.88Dec7153323@triceratops.cis.ohio-state.edu> karl@triceratops.cis.ohio-state.edu (Karl Kleinpaste) writes: =bill@ssbn.WLK.COM (Bill Kennedy) writes: = There's another thing that would help a lot. If the leaves only take groups = that people actually read then that amount of traffic is eliminated... = = My newsfeed has awarded "the most obfuscated sys line" prize... = = ...If we started at the leaves (ssbn is = really a leaf even though we feed a couple of sites) and made up the = "obfuscated sys line" and sent it up, then our feeds could consolidate them = and cut out the groups they don't need to carry. = =While I think the idea is very good, in practice it causes other =problems of its own. Notable, from my perspective, is the trouble =caused by rnews spending lots of time parsing the sys file. Our =limited resource is mostly CPU (and phone lines; 10 isn't enough for =the number of news neighbors and anon uucp'ers we've got), and I get =really nervous when I see osu-cis unbatching simultaneously from all =of att, killer, and tut.cis.ohio-state.edu, all while news is being =rebatched with compression to the leaf sites, because the CPU is being =flogged into the ground - load, as measured from a utility similar to =BSD's `uptime,' shows numbers in the vicinity of 11-14 in such cases. =Rnews takes over the system. = =We actually had to request that one particular site, which had an =extremely convoluted sys line on osu-cis, to move on and take a feed =elsewhere around town; his sys line was huge, but contained exactly =the set of groups he wanted. The tradeoff was not acceptable. The =removal of that one site (18 others remain) was enough to leave the =system responding acceptably. Reducing the amount of news sent didn't =increase the ability of my system to feed news. = =It's more efficient for massfeed sites to have all downstream feeds be =relatively simple in terms of distribution, as opposed to worrying =about the Mbyte volume going over the lines. = =--Karl Karl, Is there a way to measure where the breakpoint is? I have leafs running 20 MB Unix PCs and there's no way they can take a full feed. Pete -- Pete Holsberg UUCP: {...!rutgers!}princeton!mccc!pjh Mercer College CompuServe: 70240,334 1200 Old Trenton Road GEnie: PJHOLSBERG Trenton, NJ 08690 Voice: 1-609-586-4800
sl@van-bc.UUCP (pri=-10 Stuart Lynne) (12/10/88)
In article <486@mccc.UUCP> pjh@mccc.UUCP (Pete Holsberg) writes: >In article <KARL.88Dec7153323@triceratops.cis.ohio-state.edu> karl@triceratops.cis.ohio-state.edu (Karl Kleinpaste) writes: >=bill@ssbn.WLK.COM (Bill Kennedy) writes: >=It's more efficient for massfeed sites to have all downstream feeds be >=relatively simple in terms of distribution, as opposed to worrying >=about the Mbyte volume going over the lines. > Is there a way to measure where the breakpoint is? I have leafs >running 20 MB Unix PCs and there's no way they can take a full feed. > I've been doing much the same for the last two years. Basically I balance off the cost of carrying and producing extra batch files against modem time. I've told some sites with 1200 bps modems they *couldn't* have a full feed because we just don't have enough modem time to feed them, and instead give them a subset (usually all news/comp/sci plus rec.humor etc). I've told other small sites with a Telebits that they get a full feed and can throw away what they don't need at their end. I don't want to have to batch anymore than I have to so just give them copies of already batched files. -- Stuart.Lynne@wimsey.bc.ca {ubc-cs,uunet}!van-bc!sl Vancouver,BC,604-937-7532
cl@datlog.co.uk (Charles Lambert) (12/10/88)
Patrick Townson (ptownson@bu-cs.BU.EDU, article 26469@bu-cs.BU.EDU) rightly prods our consciences when he reminds us how needlessly costly it is to quote, verbatim, huge tracts of articles we respond to. It's a lazy habit. Magazine correspondents manage quite well without it, carrying on long threads of discussion that may be punctuated by intervals of weeks or months. We only have to remember the substance of a discussion for a few minutes or a couple of days at the worst. A brief paraphrase, preferably carried inside a substantive comment, is all that's needed. For example: in the first sentence of this posting I believe I have established the context of this response in a 15 word paraphrase (plus the formal reference, for those who want the original text). And I've stated my feelings about it in the same sentence. It may not be Mark Twain, but it's economical. Mostly, we use the included text because the followup function delivers it on a plate. I remember using "readnews" when it didn't have that built-in convenience; consequently, almost never included quotes. Perhaps cost- conscious news administrators should modify all their news reading software so that it doesn't give the prepacked inclusion, or even an environment reference to it (like $A in readnews). If we really have to try to pull in a quote, we'll probably think harder and then use a short manual quote. Now I should admit that I've got the automatically included text here while I write this. It's a handy memory jogger and I'm deleting each point as I respond to it; but I'd get by without it. Patrick writes: > Why not try writing creative replies >incorporating as little as required of other people's text right in your >own message? You're right, I didn't really need that quote but I couldn't put it better myself. Charles Lambert cl@datlog.co.uk LONDON, UK P.S. Instead of feeding us second-hand text, perhaps the news readers should put postings through spell(1) before accepting them - there are far too many "responces" on this network.
RWC102@PSUVM (R. W. F. Clark) (12/10/88)
In article <1988Dec8.172749.11645@utzoo.uucp>, henry@utzoo.uucp (Henry Spencer) says: > >Unfortunately, signatures containing network address, not just name, are >a practical necessity if you want people to be able to get mail to you. >The article-header information is *not* sufficient except in very favorable >cases. > Actually, I can usually get r to work about a third of the time. In about seventy-five percent of the other cases, the Path: line will be sufficient. It seems that only in the more unfavorable cases is the information not sufficient. fc
ahby@bungia.Bungia.MN.ORG (Shane P. McCarron) (12/12/88)
In article <2013@van-bc.UUCP> sl@van-bc.UUCP (pri=-10 Stuart Lynne) writes:
A lot of very lucid stuff, but laments that he must turn down sites and/or
waste bandwidth just because he doesn't want to batch up more news,
and waste more time and spool space.
It seems like it is time to talk about multicast again. News 2.11 has
a facility in it called multicast. While this goes largely
unutilized, it can be used to ELIMINATE wasted spool space and CPU
time. I have tried several times to post the software associated with
using this to comp.sources.unix, but Rich never seems to get around to
it.
The software takes batch files created with the M type keyword and
builds logical subsets of news based on subscription requirements.
These subsets are then batched and queued via a multi-queueing system
called uuast. Other methods would also work (uux -l, for instance).
The upshot is that for every article that comes into your system and
needs to go out again, compress is run once, batch is run once, and it
uses one unit of disk space. This is a real win for any node which
feeds news to more than one downstream site.
If there is sufficient interest in this stuff, I will try once again
to post it. Or maybe post it to comp.sources.misc?
--
Shane P. McCarron UUCP: ahby@bungia.mn.org
Systems Analyst ATT: +1 612 224-9239eric@snark.UUCP (EricS.Raymond) (12/13/88)
In article <581@bungia.bungia.mn.org>, Shane P. McCarron writes: > It seems like it is time to talk about multicast again. News 2.11 has > a facility in it called multicast. While this goes largely > unutilized, it can be used to ELIMINATE wasted spool space and CPU > time. I have tried several times to post the software associated with > using this to comp.sources.unix, but Rich never seems to get around to > it. Shene's multicast facility is fully supported and documented in 3.0. -- Eric S. Raymond (the mad mastermind of TMN-Netnews) Email: eric@snark.uu.net CompuServe: [72037,2306] Post: 22 S. Warren Avenue, Malvern, PA 19355 Phone: (215)-296-5718
friedl@vsi.COM (Stephen J. Friedl) (12/14/88)
In article <1988Dec8.172749.11645@utzoo.uucp>, henry@utzoo.uucp (Henry Spencer) writes: > Including physical address and phone number is > pointless, so there's no need to clutter up the > signature with them. Not necessarily. I include my phone number in my .sig because people *do* call me with questions, and if somebody posts a hot problem and I know the answer, I call if I can. Steve -- Stephen J. Friedl 3B2-kind-of-guy friedl@vsi.com V-Systems, Inc. attmail!vsi!friedl Santa Ana, CA USA +1 714 545 6442 {backbones}!vsi!friedl Nancy Reagan on my new '89 Mustang GT Convertible: "Just say WOW!"
jerry@olivey.olivetti.com (Jerry Aguirre) (12/15/88)
In article <952@dlhpedg.co.uk> cl@datlog.co.uk (Charles Lambert) writes: >Mostly, we use the included text because the followup function delivers it >on a plate. I remember using "readnews" when it didn't have that built-in >convenience; consequently, almost never included quotes. Perhaps cost- >conscious news administrators should modify all their news reading software >so that it doesn't give the prepacked inclusion, or even an environment Or how about the best of both worlds. Have the followup function insert the included text but flag it with something other than ">". (Except for the attribute.) Say it gets flagged with "@>" or something. Then delete those lines from the text before posting. This would mean that posters would have to manually select those lines that they wanted to include and change the prefix string. (For example delete the @ at the beginning of the line.) Yet they would still have the full text to refer to while composing their response.
night@pawl6.pawl.rpi.edu (Trip Martin) (12/20/88)
In article <952@dlhpedg.co.uk> cl@datlog.co.uk (Charles Lambert),
and Patrick Townson (ptownson@bu-cs.BU.EDU, article 26469@bu-cs.BU.EDU):
Charles Lambert takes up Patrick Townson's point about included quotes in
news articles and suggests that news administrators disable this feature.
How about letting the news software provide the solution, since it's part
of the problem. Since the references are already included with articles,
why not give news readers an option to pull up the reference articles?
The only problem I would see with this is some sites have very short expire
times, which would make this feature useless.
--
Trip Martin
night@pawl.rpi.edu
night@paraguay.acm.rpi.edujep@fantasci.UUCP (Joseph E Poplawski) (12/20/88)
In article <968@vsi.COM> friedl@vsi.COM (Stephen J. Friedl) writes: >In article <1988Dec8.172749.11645@utzoo.uucp>, henry@utzoo.uucp (Henry Spencer) writes: >> Including physical address and phone number is >> pointless, so there's no need to clutter up the >> signature with them. > >Not necessarily. I include my phone number in my .sig because >people *do* call me with questions, and if somebody posts a hot >problem and I know the answer, I call if I can. And I have had several instances where I have been called at home. One instanceI had sent someone some source code that they posted they needed, and they then called me at home to thank me (more personal than mail). In another, a question I had asked which needed a software solution helped me receive USmail from a few different software vendors and even followup calls. So I think that anyone who wishes to include their address/telephone # in their signature should do so. Just be prepared for some "junk" mail to start showing up in your USmail box... -Jo ------------------------------------------------------------------------------- | Joseph E Poplawski (Jo) US Mail: 1621 Jackson Street | | Cinnaminson NJ 08077 | | UUCP:..!rutgers!rochester!moscom!telesci!fantasci!jep | | ..!princeton!telesci!fantasci!jep | | ..!pyrnj!telesci!fantasci!jep Phone: +1 609 786-8099 home | -------------------------------------------------------------------------------
guy@auspex.UUCP (Guy Harris) (12/21/88)
>The only problem I would see with this is some sites have very short expire >times, which would make this feature useless. The other problem is that the referencing article can (and, at times, does) arrive before the referenced article (assuming the referenced article arrives at all). Netnews articles are datagrams.
eric@snark.UUCP (Eric S. Raymond) (12/23/88)
In article <766@auspex.uucp>, guy@auspex.UUCP (Guy Harris) writes: > The other problem is that the referencing article can (and, at times, > does) arrive before the referenced article (assuming the referenced > article arrives at all). Netnews articles are datagrams. Too true. I am working on a solution to this problem. The first part, already implemented in beta 3.0, stores the reference IDs of 'orphaned' child articles in special stub entries in the history file. Thus when the parent article arrives, the orphan IDs can be filled into a 'Back-Reference' header for conversation-following. The second part, not yet implemented, will enhance ihave/sendme for a true distributed fetch capability, allowing references to a missing parent to be optionally translated to a sendme request to some set of 'close neighbors' presumably connected over a high-speed link. In particular, the Internet would become a set of 'close neighbors', permitting the news volumes of the entire Internet to be accessed as a self-organizing distributed database. The step after that, of course, is 'HyperNews' -- full distributed hypertext... -- Eric S. Raymond (the mad mastermind of TMN-Netnews) Email: eric@snark.uu.net CompuServe: [72037,2306] Post: 22 S. Warren Avenue, Malvern, PA 19355 Phone: (215)-296-5718