kre@munnari.OZ (Robert Elz) (05/09/86)
In article <6657@utzoo.UUCP>, henry@utzoo.UUCP (Henry Spencer) writes: > Clearly, then, a Real Network has to have connections live all the time so > that it knows when machines go down. This isn't necessary or sufficient. The only way that (I am aware of) to determine when a machine goes down is for it to tell you that it is going down. For this, decicated connections will do, as will dial up connections. It the host goes down without telling you in advance, then all you can ever determine is that you can't communicate with it when you want to, you might surmise that its down, but there's no way you can know. Now, with either dedicated or dial up links, you can make this assumption whenever its appropriate - on a dedicated link it might be appropriate after a few seconds or minutes. On a dial up link, it might be appropriate after a few hours or days. In either case, a Real Network can then reroute traffic after it has made this determination. Robert Elz kre%munnari.oz@seismo.css.gov seismo!munnari!kre
henry@utzoo.UUCP (Henry Spencer) (05/13/86)
> If the host goes down without telling you in advance, then all you > can ever determine is that you can't communicate with it when you > want to... On a dial up link, [deciding he's down] might be appropriate > after a few hours or days. > > In either case, a Real Network can then reroute traffic after it > has made this determination. Unfortunately, such rerouting attempts are of little use (and may even be counterproductive) if the he's-down decision process takes a length of time comparable to the probable downtime. Which can easily be the case, given heavily-used phone lines and calling-time restrictions. By the time you are sure he is down, he's probably up again. Possibly things are better in Australia, but hereabouts it is impossibly hard to automatically keep down/up data current enough to be very useful for rerouting. Oh sure, if he's down for a month that's different... but it's also very infrequent, to the point where it can be handled manually. -- Join STRAW: the Society To Henry Spencer @ U of Toronto Zoology Revile Ada Wholeheartedly {allegra,ihnp4,decvax,pyramid}!utzoo!henry
honey@down.FUN (Peter Honeyman) (05/14/86)
henry points out difficulties in rerouting when the time to decide a neighbor is down is of the same order as the neighbor's downtime. to amplify, it is very easy to concoct scenarios in which cycles are created. and if steps are taken to avoid them, traffic might get blocked permanently. as henry points out, the downtime decision has to have a long period; i would guess a day or so. but then, honey danber returns queued mail after a day or so, giving the sender the option to resend along a different path. peter
zben@umd5.UUCP (Ben Cranston) (05/16/86)
In article <6657@utzoo.UUCP> henry@utzoo.UUCP (Henry Spencer) writes: > In article <?> ? writes: >> ... Real Networks do this [routing] automatically and dynamically... >Clearly, then, a Real Network has to have connections live all the time so >that it knows when machines go down. Do let us know when you are willing >to finance such a network for us. Meanwhile, we will continue to use our >unReal but still exceedingly cost-effective network. We will also continue >to have a low opinion of people who think that there is something "unreal" >about one of the largest networks in the world. Ahem... In the finest tradition of "layered systems" UUCP is built ON TOP OF a "REAL SYSTEM". I am speaking of the international telephone system. So when I send mail from UMD2 I open an SMTP connection; when you send mail you open a TELCO connection. We both rely upon the underlying transport layer to do the dirty work. Yes, Telco is live all the time. Yes, users VERY seldom know when a piece of Telco is down. It's a pretty fine network, all in all, and sometimes I wonder why we bother to reinvent the wheel. The only reason UUCP need use multiple systems is to try to take advantage of things like "unlimited local calling". If it were not for these quite artificial constraints (i.e. if telco charged for straight "mile-seconds") it would be just as economcal to call the final site as to chain through intermediate sites. This is of course ignoring the site-phone# database problem... I personally have a very low opinion of people who thing bigger is better. If you dislike those of us in the "lower 48" claiming we are better because we are bigger, you might want to rethink your statement... -- "We're taught to cherish what we have | Ben Cranston by what we have no longer..." | zben@umd2.umd.edu ...{seismo!umcp-cs,ihnp4!rlgvax}!cvl!umd5!zben
avolio@decuac.UUCP (05/16/86)
In article <961@umd5.UUCP>, zben@umd5.UUCP (Ben Cranston) writes: > In the finest tradition of "layered systems" UUCP is built ON TOP OF > a "REAL SYSTEM". I am speaking of the international telephone system. Look. I am sorry I used the term Real Network, okay? I didn't mean to besmirch UUCP, Ma Bell (wherever she is...), or anyone or anything. I love UUCP. It is wonderful to get a call from someone who wants a net connection (see I even call it a net!) and to have it up and running in 10 minutes. No kidding, I think it is tops. (As opposed to TOP, which is something different.) The distinction was drawn wrt how routing is done or figured out. Yes, the phone call from here to there is routed automatically with no noticable loss of signal and it is done completely transparently to the user. But, I was referring to the routing that is done on top of the phone lines. One has to know the route from one machine to another or at least to a "smarter" UUCP host. Then, in path a!b!c!d!user if host c is down, there is nothing automatic to retoute the mail message via another route. That is the short coming that was trying to bring out ( ... at least I think it was ... it's been a few weeks). -- Fred @ DEC Ultrix Applications Center INET: avolio@decuac.DEC.COM * Fight the Fight * UUCP: {decvax,seismo,cbosgd}!decuac!avolio * Rescue the Unborn *
henry@utzoo.UUCP (Henry Spencer) (05/22/86)
> >... We will also continue > >to have a low opinion of people who think that there is something "unreal" > >about one of the largest networks in the world. > > I personally have a very low opinion of people who think bigger is better. Just as well that I don't think that, then. Please read my words more carefully. If there is an elephant on your doorstep, it is foolish to ignore it, even if you happen to feel that it is ugly. Some of us get very tired of if-it's-not-like-the-Arpanet-then-it's-beneath-our-notice. -- Join STRAW: the Society To Henry Spencer @ U of Toronto Zoology Revile Ada Wholeheartedly {allegra,ihnp4,decvax,pyramid}!utzoo!henry