jbuck@epimass.EPI.COM (Joe Buck) (09/21/88)
Sites using the pathalias data generate horrible paths to Stanford
University, among other places. The reason is that Stanford is on the
Internet, but one poorly-connected Stanford site is listed as the only
gateway to .stanford.edu in the UUCP map data. Here is their entry:
#N wso
#S DEC VAX-11/780; 4.3 BSD UNIX
#O Wilcox Solar Observatory, Stanford University
#C Rick Bogart
#E wso!rick
#T +1 415 723 1296
#P CSSA, Stanford, CA 94305-4055
#L 37 24 35 N / 122 10 02 W
#R Wilcox Solar Observatory at Stanford University; position +/- 2"
#U
#W wso!rick (Rick Bogart); Wed Jul 13 14:37:04 PDT 1988
#
wso .stanford.edu
wso= wso.stanford.edu
wso mtxinu(DIRECT), ncar(DEMAND), noao(DEMAND)
If this map entry were not there, a site that talks directly to uunet,
when mailing to luser@foo.stanford.edu, would search, in sequence, for
luser.stanford.edu (not found)
.stanford.edu (not found)
.edu (found, gateway is uunet)
and would use the path
uunet!foo.stanford.edu!luser
and uunet would ship the message via the Internet to Stanford.
Instead, horrible paths are generated. In my case, I get the path
oliveb!ames!ncar!noao!wso!foo.stanford.edu!luser
Gag! Awful. At minimum, the line
wso .stanford.edu
should be deleted from the map entry. Possibly the whole map entry
should be deleted (at least after the half-price Trailblazers
arrive! :-).
There should be NO gateway listed to second-level Internet domains
in cases where the top-level domain servers will get mail through just
as fast. Let's keep the maps small and consider what the pathalias
results will be when we add information.
--
- Joe Buck, card-carrying ACLU liberal
jbuck@epimass.epi.com, or uunet!epimass.epi.com!jbuck,
or jbuck%epimass.epi.com@uunet.uu.net for old Arpa siteswerner@utastro.UUCP (Werner Uhrig) (09/21/88)
I will forward the article to the postmaster at wso.stanford.edu with a request to submit a updated map-entry; I also wonder if/why-not regional map-administrators do any sanity-checks on submitted map-entries and if they should not take uni-lateral corrective action when such a problem is reported?! BTW, I have the feeling that several similar "problematic" map-entries exist; one must be in the vicinity of mimsy, as I have seen some crazy (and failing) paths routed through there ... -- --------------------> PREFERED-RETURN-ADDRESS-FOLLOWS <--------------------- (ARPA) werner@rascal.ics.utexas.edu (Internet: 128.83.144.1) (INTERNET) werner%rascal.ics.utexas.edu@cs.utexas.edu (UUCP) ..!utastro!werner or ..!uunet!rascal.ics.utexas.edu!werner
vixie@decwrl.dec.com (Paul Vixie) (09/21/88)
# There should be NO gateway listed to second-level Internet domains # in cases where the top-level domain servers will get mail through just # as fast. While I agree that there are problems caused by it, this will make things even worse. Consider decwrl, known to pathalias as the gateway to .dec.com. Consider vixie.sf.ca.us, which speaks directly to decwrl and also to uunet. Consider mail send from paul@vixie.sf.ca.us to reid@decwrl.dec.com. Without the "decwrl .dec.com" entry in decwrl's map entry, the mail would go to uunet to get to decwrl, even though decwrl is a neighbor. Same is true for all .dec.com sites: decwrl really is the fastest way to get there for many UUCP sites. It can be solved by having all neighbors add "decwrl .dec.com" to their /usr/lib/mail/glue.local files, but this is both irritating and incomplete: incomplete because the problem also exists for many non-neighbors, though to a lesser degree. So, again: I agree that there is a problem; this isn't the way to solve it. The whole method of naming directly-connected Internet domains in pathalias map entries needs rethinking. I have no wonderful ideas, though. Discussion? -- Paul Vixie Work: vixie@decwrl.dec.com decwrl!vixie +1 415 853 6600 Play: paul@vixie.sf.ca.us vixie!paul +1 415 864 7013
jerry@olivey.olivetti.com (Jerry Aguirre) (09/22/88)
In article <2488@epimass.EPI.COM> jbuck@epimass.EPI.COM (Joe Buck) writes: >Sites using the pathalias data generate horrible paths to Stanford >University, among other places. The reason is that Stanford is on the >Internet, but one poorly-connected Stanford site is listed as the only >gateway to .stanford.edu in the UUCP map data. Here is their entry: >wso .stanford.edu >wso= wso.stanford.edu >wso mtxinu(DIRECT), ncar(DEMAND), noao(DEMAND) What's the matter? You don't like sending mail to Stanford via Colorado and Arizona? :-) Actually it is worse than that pathalias data would indicate. It is only used for paths FROM wso to those sites. (Otherwise your mail would have been sent via mtxinu.) Paths going to wso must use the noao-wso link because it is the ONLY link to wso listed in the maps. So, in this case, badly connected means just one (1) incomming connection. If they had a few more connections or even if mtxinu would show their return "DIRECT" connection then it wouldn't be so bad. Jerry Aguirre
emv@mailrus.cc.umich.edu (Edward Vielmetti) (09/22/88)
In article <29499@oliveb.olivetti.com> jerry@olivey.UUCP (Jerry Aguirre) writes: >In article <2488@epimass.EPI.COM> jbuck@epimass.EPI.COM (Joe Buck) writes: >>Sites using the pathalias data generate horrible paths to Stanford >>University, among other places. The reason is that Stanford is on the >>Internet, but one poorly-connected Stanford site is listed as the only >What's the matter? You don't like sending mail to Stanford via Colorado >and Arizona? :-) pathalias data is the wrong source of information for .edu domain hosts; the name servers are much more useful and reliable. filter out all of the .foo.edu information from your pathalias run and just make sure that the .edu gateway you declare is running a sendmail with MX records. .gov, .mil, and .net should be handled similarly. .com might be best left in the maps depending on the local density of .com sites not directly on the internet. I'm collecting pathalias post-processor filters - mail them to emv@mailrus.cc.umich.edu and I'll summarize and re-post. --Ed
grandi@noao.edu (Steve Grandi) (09/23/88)
In article <29499@oliveb.olivetti.com> jerry@olivey.UUCP (Jerry Aguirre) writes: >In article <2488@epimass.EPI.COM> jbuck@epimass.EPI.COM (Joe Buck) writes: >>Sites using the pathalias data generate horrible paths to Stanford >>University, among other places. > >What's the matter? You don't like sending mail to Stanford via Colorado >and Arizona? :-) > Hmm; no wonder I have all this traffic to wso! Things are particularly clogged up right now since wso isn't answering the phone. I've manually gone through the messages waiting for wso and revectored those meant for someone@somewhere.stanford.edu via the Internet. I've put a hack in my sendmail.cf so that mail sent to wso!string will now be sent to string@wso.stanford.edu so messages should now bounce much faster! An amusing note: Rick Bogart, the postmaster at wso, had sent a map update which no longer had the offending gateway line to cbosgd!uucpmap which got bounced to my machine awaiting a connection to wso! I revectored this message to uucpmap@rutgers.edu... -- Steve Grandi, National Optical Astronomy Observatories, Tucson AZ, 602-325-9228 UUCP: {arizona,decvax,ncar}!noao!grandi or uunet!noao.edu!grandi Internet: grandi@noao.edu SPAN/HEPNET: NOAO::GRANDI (NOAO=5355)
aad@stpstn.UUCP (Anthony A. Datri) (09/24/88)
>pathalias data is the wrong source of information for .edu domain >hosts; the name servers are much more useful and reliable. >filter out all of the .foo.edu information from your pathalias >run and just make sure that the .edu gateway you declare is running >a sendmail with MX records. If I'm within a couple hops of a uucp-only .edu domain, sending mail for it across the net to (for example) uunet and uunet sending it back seems like a real waste of time and links. It's going to have to go across uucp links anyway -- why bother the gateway sites with the traffic and why bother the arpanet with the traffic? -- @disclaimer(Any concepts or opinions above are entirely mine, not those of my employer, my GIGI, or my 11/34) beak is beak is not Anthony A. Datri,SysAdmin,StepstoneCorporation,stpstn!aad
jbuck@epimass.EPI.COM (Joe Buck) (09/27/88)
In article <697@mailrus.cc.umich.edu> emv@mailrus.cc.umich.edu (Edward Vielmetti) writes: >pathalias data is the wrong source of information for .edu domain >hosts; the name servers are much more useful and reliable. >filter out all of the .foo.edu information from your pathalias >run and just make sure that the .edu gateway you declare is running >a sendmail with MX records. Correction: pathalias data is the wrong source of information for .edu, .com, .mil, etc, domains IF the domain is connected to the Internet or to CSNET, unless the gateway is also very well connected to major sites on the UUCP network (as Sun is, for example). Otherwise (for domains that are connected to the outside world only by UUCP) it is frequently more efficient to use pathalias data, since such data will show all connections. You don't necessarily know that a particular domain is UUCP-only or not. There has been some controversy lately about whether to show domain information for non-UUCP Zone members in the d.* files. There's a good reason not to; it generates WORSE mail paths in many cases if it's not done correctly. But don't go blindly removing all second level domain paths from your pathalias output (for example, don't remove .epi.com!). Mail will still get through, but everything will pass through the top-level domain gateways whether it needs to or not, increasing the load on sites that are already extremely busy. -- - Joe Buck, card-carrying ACLU liberal jbuck@epimass.epi.com, or uunet!epimass.epi.com!jbuck, or jbuck%epimass.epi.com@uunet.uu.net for old Arpa sites
wcf@psuhcx.psu.edu (Bill Fenner) (09/28/88)
In article <2140@stpstn.UUCP> aad@stpstn.UUCP (Anthony A. Datri) writes: |>pathalias data is the wrong source of information for .edu domain |>hosts; the name servers are much more useful and reliable. |>filter out all of the .foo.edu information from your pathalias |>run and just make sure that the .edu gateway you declare is running |>a sendmail with MX records. | |If I'm within a couple hops of a uucp-only .edu domain, sending mail |for it across the net to (for example) uunet and uunet sending it back |seems like a real waste of time and links. It's going to have to go |across uucp links anyway -- why bother the gateway sites with the |traffic and why bother the arpanet with the traffic? | How about this- pathalias checks the cost of a second-level domain path, compares it to the cost of the first-level path. If it's cheaper, then the second-level path is put in the database. If not, it'll put it in the first-level path. If they're equal, it'll put in the second-level path. Bill -- Bitnet: wcf@psuhcx.bitnet Bill Fenner | "Ain't got no cash, Internet: wcf@hcx.psu.edu | Ain't got no style UUCP: {gatech,rutgers}!psuvax1!psuhcx!wcf | Ain't got no girls Fido: Sysop at 263/42 (814/238 9633) \hogbbs!wcf| To make me smile"