zeeff@b-tech.UUCP (Jon Zeeff) (08/30/88)
Perhaps all this discussion can result in some conclusions on a few basic questions. As it is, each administrator seems to sit back and come up with what he thinks is right. The rfc's are either ignored or don't seem to address the problems people have. It might help if we had something we could point to and say "95% of sites disagree with you on x - perhaps you should conform". 1) Should active rerouting be done? (ie, should a xxx!yyy!zzz route be changed to www!yyy!zzz if the maps that a site has say it is cheaper) 2) Should a 100% correct rfc 822 compatible From: line be modified if it is just passing through a site? Assume registered domain names for all hosts involved. 3) If yes to #2, what format should be used to rewrite "From: user@site.com": a) "From: user%site.com@thissite.edu" b) "From: @thissite.edu:user@site.com". c) other 3) Should a internet to uucp gateway rewrite From: lines into ! format? 4) What should a uucp to internet gateway do with From: lines of the format "From: user@site.uucp". a) leave them alone b) Change to "From: user%site.uucp@thissite.com" c) Change to "From: site!user@thissite.com" d) other 5) What should a uucp to internet gateway do with From: lines of the form site1!site2!user? a) leave them alone b) Change to "From: user%site2%site1@thissite.com" c) Change to "From: site1!site2!user@thissite.com" d) other 6) Should the address "zzz!user%site.com" be sent to site.com or zzz? (ie, what should have precedence) 7) Should sites with registered domain names put these names in the uucp maps? Mail your results to me. Additional questions are welcome. -- Jon Zeeff Branch Technology, umix!b-tech!zeeff zeeff%b-tech.uucp@umix.cc.umich.edu
zeeff@b-tech.UUCP (Jon Zeeff) (09/04/88)
I received 27 responses to this survey. A few responses were difficult to interpret, but the results are roughly: 1) Should active rerouting be done? (ie, should a xxx!yyy!zzz route be changed to www!yyy!zzz if the maps that a site has say it is cheaper) 79% No, 14% Yes, 7% Maybe 2) Should a 100% correct rfc 822 compatible From: line be modified if it is just passing through a site? Assume registered domain names for all hosts involved. 96% No, 4 % Maybe 3) Should a internet to uucp gateway rewrite From: lines into ! format? 56% No, 36% Yes, 8% Maybe 4) What should a uucp to internet gateway do with From: lines of the format "From: user@site.uucp". a) leave them alone 24% b) Change to "From: user%site.uucp@thissite.com" 21% c) Change to "From: site!user@thissite.com" 41% d) other 10% 5) What should a uucp to internet gateway do with From: lines of the form site1!site2!user? a) leave them alone 16% b) Change to "From: user%site2%site1@thissite.com" 24% c) Change to "From: site1!site2!user@thissite.com" 48% d) other 12% 6) Should the address "zzz!user%site.com" be sent to site.com or zzz? (ie, what should have precedence) 57% for zzz, 22% for site.com 7) Should sites with registered domain names put these names in the uucp maps? 88% Yes, 4% No, 8% Maybe My analysis (for what it's worth) is in another article. Raw data is available on request. -- Jon Zeeff Branch Technology, umix!b-tech!zeeff zeeff%b-tech.uucp@umix.cc.umich.edu -- Jon Zeeff Branch Technology, umix!b-tech!zeeff zeeff%b-tech.uucp@umix.cc.umich.edu
zeeff@b-tech.UUCP (Jon Zeeff) (09/04/88)
> > 1) Should active rerouting be done? (ie, should a xxx!yyy!zzz route be > changed to www!yyy!zzz if the maps that a site has say it is cheaper) > > 79% No, 14% Yes, 7% Maybe > Clearly the majority does not want active rerouting. Sites that want to do so should keep things working by marking all links as DEAD in the map entrys that they circulate. > 2) Should a 100% correct rfc 822 compatible From: line be modified if The conclusion is that From lines should not be touched if they are ok (ie, certainly don't create something like "user%origin.com@thissite.edu" from "user@origin.com"). If there are unregistered uucp sites, convert to the form: From: site2!site!user@thissite.com (even if rfc976 suggests you don't) > 6) Should the address "zzz!user%site.com" be sent to site.com or zzz? > (ie, what should have precedence) Most people say zzz so let's standardize on that (and avoid creating things like this whenever possible). > 7) Should sites with registered domain names put these names in the uucp > maps? > 88% said yes. The no answers mentioned that doing so might cause mail to be routed completely via uucp (vs. being routed to the nearest internet site), which might slow things down. -- Jon Zeeff Branch Technology, umix!b-tech!zeeff zeeff%b-tech.uucp@umix.cc.umich.edu
gmp@rayssd.ray.com (Gregory M. Paris) (09/04/88)
In article <4748@b-tech.UUCP> zeeff@b-tech.UUCP (Jon Zeeff) writes: > > > > 1) Should active rerouting be done? (ie, should a xxx!yyy!zzz route be > > changed to www!yyy!zzz if the maps that a site has say it is cheaper) > > > > 79% No, 14% Yes, 7% Maybe > > > > Clearly the majority does not want active rerouting. Sites that want > to do so should keep things working by marking all links as DEAD in > the map entrys that they circulate. I don't understand how marking all links as DEAD helps. Presumably, a site that's doing active rerouting is using pathalias data to do so. Mail traveling through the rerouting site does so because: 1. The originating site routed the mail using pathalias data, in which case the route probably will not be altered by the rerouting site. DEAD has no beneficial effect. 2. The mail was routed "manually" by a human that rightly or wrongly knows a good route. DEAD has no effect whatsoever on manual routing. Explanation? -- Greg Paris <gmp@rayssd.ray.com> {decuac,gatech,necntc,sun,uiucdcs,ukma}!rayssd!gmp NO KILL I
zeeff@b-tech.UUCP (Jon Zeeff) (09/04/88)
In article <3703@rayssd.ray.com> gmp@rayssd.RAY.COM (Gregory M. Paris) writes: >In article <4748@b-tech.UUCP> zeeff@b-tech.UUCP (Jon Zeeff) writes: >> >> Clearly the majority does not want active rerouting. Sites that want >> to do so should keep things working by marking all links as DEAD in >> the map entrys that they circulate. > >I don't understand how marking all links as DEAD helps. Presumably, Mail that was completely routed by up to date, consistent pathalias data has never been a problem. One problem is created when a site thinks that it has better map info than the originating site. Marking links as DEAD avoids these sites and this problem. Another problem (which I don't have much sympathy for) is duplicate site names. Many people partially route an address by using site1!site2!user and letting their mailer figure out how to get to site1. Marking links as DEAD avoids sites that will try to take short cuts to site2. > 2. The mail was routed "manually" by a human that rightly or > wrongly knows a good route. DEAD has no effect whatsoever By marking the links as DEAD, you have at least documented the problem. If someone ignores this information and uses completely manual routing through nasty sites, they get what they asked for. -- Jon Zeeff Branch Technology, umix!b-tech!zeeff zeeff%b-tech.uucp@umix.cc.umich.edu
gmp@rayssd.ray.com (Gregory M. Paris) (09/05/88)
In reply to my observation that active rerouters using pathalias data have no effect on mail routed using pathalias (thus marking links as DEAD has no beneficial effect), in article <4750@b-tech.UUCP> zeeff@b-tech (Jon Zeeff) writes: > Mail that was completely routed by up to date, consistent pathalias > data has never been a problem. One problem is created when a site > thinks that it has better map info than the originating site. Marking > links as DEAD avoids these sites and this problem. Another problem I take from this that you are saying that it is OK to actively reroute as long as you use up to date, consistent pathalias data. I don't see where your poll came to that conclusion. In fact, I'd say that you are DEAD (pun) wrong here, as most complaints on this subject have been about rutgers, which supposedly has the most up to date pathalias info of any site. The results of your poll were interesting, but I still don't think your conclusion that active rerouters should mark all links as DEAD makes sense. I also wonder what kind of enforcement mechanism you are going to back up this directive (suggestion?, plea?) with. Site admins that reroute probably do it for what they think are good reasos. What's going to make them do what you and/or the majority want them to do? -- Greg Paris <gmp@rayssd.ray.com> {decuac,gatech,necntc,sun,uiucdcs,ukma}!rayssd!gmp NO KILL I
jos@idca.tds.PHILIPS.nl (Jos Vos) (09/05/88)
In article <4748@b-tech.UUCP> zeeff@b-tech.UUCP (Jon Zeeff) writes:
]
]The conclusion is that From lines should not be touched if they are ok (ie,
]certainly don't create something like "user%origin.com@thissite.edu"
]from "user@origin.com"). If there are unregistered uucp sites, convert
]to the form:
]
]From: site2!site!user@thissite.com
](even if rfc976 suggests you don't)
This means that we assume that everybody agrees on the precedence...
Did you know that even in documentation of mailers like Elm the
precedence is wrong? (I didn't check the code...).
Or am I wrong?
]> 6) Should the address "zzz!user%site.com" be sent to site.com or zzz?
]> (ie, what should have precedence)
]
]Most people say zzz so let's standardize on that (and avoid creating things
]like this whenever possible).
The address zzz!user%site.com isn't a valid RFC822 address, is it?
This means that the precedence of zzz!user@site.com and zzz!user%site.com
differ?
--
-- ###### Jos Vos ###### Internet jos@idca.tds.philips.nl ######
-- ###### ###### UUCP ...!mcvax!philapd!jos ######zeeff@b-tech.UUCP (Jon Zeeff) (09/06/88)
In article <3710@rayssd.ray.com> gmp@rayssd.RAY.COM (Gregory M. Paris) writes: >In reply to my observation that active rerouters using pathalias data >have no effect on mail routed using pathalias (thus marking links as >DEAD has no beneficial effect), in article <4750@b-tech.UUCP> zeeff@b-tech >(Jon Zeeff) writes: > >> Mail that was completely routed by up to date, consistent pathalias >> data has never been a problem. One problem is created when a site >> thinks that it has better map info than the originating site. Marking >> links as DEAD avoids these sites and this problem. Another problem > >I take from this that you are saying that it is OK to actively reroute >as long as you use up to date, consistent pathalias data. I don't see >where your poll came to that conclusion. In fact, I'd say that you are >DEAD (pun) wrong here, as most complaints on this subject have been >about rutgers, which supposedly has the most up to date pathalias info >of any site. Yes, as long as you pick the exact same route I did you can reroute my mail all you want :-). Read carefully what I said - "mail that was completely routed by up to date, consistent pathalias data ...". In other words, if you can prove 1) that I used only pathalias to route my mail and 2) that you have the same map data, then yes, you should feel free to "reroute" it, because you are going to come up with exactly the same path I did. Of course no site can prove either of these things, which is why your observation in the first paragraph is incorrect (rerouters often do change routes that were created with pathalias - their map data differs). Rutgers will never have all of the most up to date map information. Local changes are known by local sites first - it seems to take rutgers awhile to get changes incorporated into the map data. > >The results of your poll were interesting, but I still don't think your >conclusion that active rerouters should mark all links as DEAD makes >sense. I guess I can only tell you to talk to others who manually mark rutgers links as DEAD. It does help. >I also wonder what kind of enforcement mechanism you are going >to back up this directive (suggestion?, plea?) with. Site admins that >reroute probably do it for what they think are good reasos. What's >going to make them do what you and/or the majority want them to do? I don't know. How would any site be handled if it advertised many low cost links and then screwed up the mail it received? Maybe a # line in the maps data marking these sites would do it. -- Jon Zeeff Branch Technology, umix!b-tech!zeeff zeeff%b-tech.uucp@umix.cc.umich.edu
lear@NET.BIO.NET (Eliot Lear) (09/06/88)
You assume that you know the better route. I claim it is not even your place to determine a route. Again, look at the havok that ihnp4 has caused those who do not actively reroute. -- Eliot Lear [lear@net.bio.net]
gmp@rayssd.ray.com (Gregory M. Paris) (09/06/88)
In article <4753@b-tech.UUCP> zeeff@b-tech.UUCP (Jon Zeeff) writes: > I guess I can only tell you to talk to others who manually mark rutgers > links as DEAD. It does help. It may help in the case of mail from sites using pathalias data different from the published data. It does not help with sites that have pathalias data that's in sync with the rerouting site's data, nor does it help if the mail has been routed manually. I submit that most sites are using the published data unaltered, and that those persons that route mail manually do not check the published map data on a regular basis (if at all). Even if you dispute these assumptions, you cannot claim that marking links as DEAD will solve the problem for everyone, so I still see it as a non-solution. More, the idea seems born of a desire to punish active rerouters, rather than improve mail service. -- Greg Paris <gmp@rayssd.ray.com> {decuac,gatech,necntc,sun,uiucdcs,ukma}!rayssd!gmp NO KILL I
willy@idca.tds.PHILIPS.nl (Willy Konijnenberg) (09/09/88)
I get the impression that some careful reading is required in this discussion. Jon and Gregory seem to consistently misinterpret eachothers texts. In article <4753@b-tech.UUCP> zeeff@b-tech.UUCP (Jon Zeeff) writes: >In article <3710@rayssd.ray.com> gmp@rayssd.RAY.COM (Gregory M. Paris) writes: >>> Mail that was completely routed by up to date, consistent pathalias >>> data has never been a problem. One problem is created when a site >>> thinks that it has better map info than the originating site. Marking >>> links as DEAD avoids these sites and this problem. >> >>I take from this that you are saying that it is OK to actively reroute >>as long as you use up to date, consistent pathalias data. Nope Jon. What Gregory says is that full routing with pathalias is not the problem, but active RErouting is. In my opinion, full routing from the very source is not entirely ok either, but that is not discussed here. >>The results of your poll were interesting, but I still don't think your >>conclusion that active rerouters should mark all links as DEAD makes >>sense. Hey, nobody said the rerouters should change their maps! What has been suggested is that if you have a problem with an active rerouter, YOU mark all links to that site as DEAD in your copy of the maps. That way, YOUR pathalias will not route THRU the offending site and your problem is solved. You don't need the cooperation of the rerouter for this. The result of this would eventually be that everybody is carefully avoiding the rerouters and the rerouters form isolated, lonesome islands in the net. -- Willy Konijnenberg Philips Telecommunication & Data Systems Domain: willy@idca.tds.PHILIPS.nl UUCP: ..!mcvax!philapd!willy