reid@decwrl.UUCP (05/22/86)
I just installed pathalias here and ran it over the newly-posted map data. I used a version that I've been running at Stanford for several months, and it seemed in the past to work fine. I asked for a path to "sdcrdcf", even though I know that the path hplabs!sdcrdcf!%s works just fine. Here is what it came up with. This, without a doubt, has to be the worst mail routing I have ever seen. It's debugging time. greipa!texsun!oz!munnari!prlb2!ukc!warwick!cvaxa!csunb!csun!sdcrdcf!%s Notes: decwrl is near San Francisco greipa was near San Francisco (it's been dead for months, though) texsun is in Texas. oz is in Washington munnari is in Australia prlb2 is in Belgium ukc is in England warwick is in England cvaxa is in England csunb is near Los Angeles sdcrdcf is near Los Angeles I'm sorely tempted to send a message over this path to see if it works, but I'm sure it won't because the "oz<-->munnari" link is a fiction: oz is the domain name for Australia.
reid@decwrl.UUCP (05/23/86)
Yesterday I said: >This, without a doubt, has to be the worst mail routing I have ever seen: >greipa!texsun!oz!munnari!prlb2!ukc!warwick!cvaxa!csunb!csun!sdcrdcf!%s Well, this was not a bug in pathalias but 3 combined bugs in the database. (1) Site "texsun" claims a link to "oz", when in fact it has a link to "ozone" on its local network. (2) Somebody replaced the carefully-commented-out links from munnari to the outside world in the entry for munnari, so that pathalias thought it could route through australia. (3) The machine named "oz", in Washington, is somehow made equivalent to the domain ".oz", which is Australia. The time cost of X.25 links is fairly low, so the pathalias entries that are made for them show very small numbers, and an article can take a large number of X.25 hops for the price of one "HOURLY" link.
salzman@rdlvax.UUCP (05/23/86)
In article <3148@decwrl.DEC.COM> reid@decwrl.UUCP (Brian Reid) writes: >I know that the path hplabs!sdcrdcf!%s works just fine. Here is what it came >up with. This, without a doubt, has to be the worst mail routing I have ever >seen. It's debugging time. > >greipa!texsun!oz!munnari!prlb2!ukc!warwick!cvaxa!csunb!csun!sdcrdcf!%s That's an impossible path!!!! You must have very buggy software or bogus maps; "csun" and "csunb" are both at Cal. State University Northridge (where I happen to attend), and "csunb" only talks to one machine - "csun". The "csun" machine is the only site at Northridge that talks to the outside world, and it does not talk to "cvaxa".... -- * Isaac Salzman (Gumby) * Research Development Labs (RDL) * Culver City, California, 90230 * UUCP: ...!{psivax,sdcrdcf,ttidca,csun}!rdlvax!salzman * ARPA: sdcrdcf!rdlvax!salzman@LOCUS.UCLA.EDU
stephen@comp.lancs.ac.uk (Stephen J. Muir) (05/24/86)
In article <3148@decwrl.DEC.COM> reid@decwrl.UUCP (Brian Reid) writes: >up with. This, without a doubt, has to be the worst mail routing I have ever >seen. It's debugging time. > >greipa!texsun!oz!munnari!prlb2!ukc!warwick!cvaxa!csunb!csun!sdcrdcf!%s Then you're running an outdated pathalias. The latest version is given the list of map files (so it knows which file each site was in). This way, it can differentiate between multiple hosts with the same name. -- UUCP: ...!seismo!mcvax!ukc!dcl-cs!stephen DARPA: stephen%comp.lancs.ac.uk@ucl-cs | Post: University of Lancaster, JANET: stephen@uk.ac.lancs.comp | Department of Computing, Phone: +44 524 65201 Ext. 4120 | Bailrigg, Lancaster, UK. Project:Alvey ECLIPSE Distribution | LA1 4YR
jhk@cs1.UUCP (05/25/86)
In article <55@rdlvax.UUCP> salzman@rdlvax.UUCP (Gumby) writes: >In article <3148@decwrl.DEC.COM> reid@decwrl.UUCP (Brian Reid) writes: >> ... This, without a doubt, has to be the worst mail routing I have ever >>seen. It's debugging time. >>greipa!texsun!oz!munnari!prlb2!ukc!warwick!cvaxa!csunb!csun!sdcrdcf!%s >That's an impossible path!!!! Yes, csunb is only hooked up directly to csun which is our central node that talks to the whole world. However, I noticed that there was a site out there somewhere (I think in Europe) that had the name "csunb". We also noticed this problem when we generated our pathalias and csunb was used to connect all over the place. Maybe we should change the name of that site? Any suggestions? -- Joe Kwan, Calif State Univ, Northridge Computer Science Department uucp: {sdcrdcf, litvax, ihnp4, hplabs, ttidca, psivax, csustan}!csun!cs1!jhk BITNET: RGFYJHK@CALSTATE ARPANET: RGFYJHK%CALSTATE.BITNET@WISCVM.WISC.EDU USMail: CSUN, Computer Science Department, Northridge, CA 91330
geoff@desint.UUCP (Geoff Kuenning) (05/27/86)
In article <185@comp.lancs.ac.uk> stephen@comp.lancs.ac.uk (Stephen J. Muir) writes: > Then you're running an outdated pathalias. The latest version is given the > list of map files (so it knows which file each site was in). This way, it > can differentiate between multiple hosts with the same name. What a kludge! I suspect there are a lot of sites out there that, for one reason or another, are forced to feed the maps to pathalias on standard input. For example, I don't have enough free space to change my current zcat u.*.Z | pathalias into uncompress u.*.Z; pathalias u.*; compress -f u.* Furthermore, this misfeature is not mentioned in any of the pathalias documentation files (I just checked). -- Geoff Kuenning {hplabs,ihnp4}!trwrb!desint!geoff
honey@down.FUN (Peter Honeyman) (05/27/86)
place the following line at the top of u.gbr.1
private {csunb} # csunb:usa.ca
and the problem goes away. here are the private definitions
that i'm currently using:
u.gbr.1 private {sol, csunb} # sol:usa.nh, csunb:usa.ca
u.usa.ca.1 private {netvax, oberon} # netvax:kor, oberon:usa.nj.att
u.usa.ca.2 private {acorn, r2d2} # acorn:gbr, r2d2:usa.co
u.usa.il.1 private {apollo} # apollo:usa.ma
u.usa.il.att.1 private {sysdes} # sysdes:gbr
u.usa.ma.1 private {saturn} # saturn:usa.oh.att
u.usa.nc.1 private {happy} # happy:usa.oh
u.usa.nh.1 private {hermes} # hermes:usa.il.att
u.usa.ny.1 private {sysdes} # sysdes:gbr
u.usa.wa.1 private {boris} # boris:usa.ny
i urge the map people to follow suit.
peterjso@edison.UUCP (John Owens) (06/04/86)
> >greipa!texsun!oz!munnari!prlb2!ukc!warwick!cvaxa!csunb!csun!sdcrdcf!%s > > That's an impossible path!!!! You must have very buggy software or bogus > maps; "csun" and "csunb" are both at Cal. State University Northridge > (where I happen to attend), and "csunb" only talks to one machine - "csun". > The "csun" machine is the only site at Northridge that talks to the outside > world, and it does not talk to "cvaxa".... Unfortunately, "cvaxa", in Great Britian, has a machine "csuna" on its local ethernet.... The biggest problem with not having a separate map entry for local nodes is that no duplicate name checking is done. This case was obvious looking at the maps; if cvaxa had been in California, it might not have been. Ideally, of course, outside mail should be routed by username by the main node.... [Even more ideally, everyone would use domains, and you wouldn't have a problem with csuna.cvaxa.UK or whatever.] Oh, well, John Owens @ General Electric Company (+1 804 978 5726) edison!jso%virginia@CSNet-Relay.ARPA [old arpa] edison!jso@virginia.EDU [w/ nameservers] jso@edison.UUCP [w/ uucp domains] {cbosgd allegra ncsu xanth}!uvacs!edison!jso [roll your own]