- Test - lists.nanog.org

Re: mail does bounce (was: Customers down?)
by Sean Donelan 03 Jan '00

03 Jan '00

On Sun, 02 January 2000, Owen DeLong wrote: > I don't even assume that anyone elses mail system works, let > alone that it works this way, or even complies with the RFC's. > > There's way too much MicroSoft stuff out there to believe > any of that. I merely purport that the above is BCP. Could you point me to a written Best Common Practice which says this? Or is it simply one of those assumptions people like to make? The problem with assumptions is they work great, until someone makes a different one.

1 0

Re: mail does bounce (was: Customers down?)
by owen＠dixon.delong.sj.ca.us 03 Jan '00

03 Jan '00

> On Sun, 02 January 2000, Owen DeLong wrote: > > I'll take a stab at this: > > > > Attempting to resolve a.b.c.d.e.f.g.h.i > > > > If you get back a return value from DNS that says > > "Authoritative answer, a.b.c.d.e.f.g.h.i does not exist" > > Bounce the mail. > > > > If you get back a return that says > > "Resolving a.b.c.d.e.f.g.h.i timed out without a return" > > Queue the mail. > > > > This also works for i, h.i, g.h.i, f.g.h.i, etc. > > I'm happy to agree this is a very sensible thing to do. However, > there is no requirement for any mailer to do it. Some mailers may, > but other mailers may not. So it is not very wise to assume all > mailers work the same as yours. > > There is a written recommendation that domains should have DNS > servers on seperate networks. There is no such recommendation > that mailers retry or queue mail when they can't reach DNS servers. > Sean, I don't even assume that anyone elses mail system works, let alone that it works this way, or even complies with the RFC's. There's way too much MicroSoft stuff out there to believe any of that. I merely purport that the above is BCP. Owen

1 0

Re: mail does bounce (was: Customers down?)
by owen＠dixon.delong.sj.ca.us 03 Jan '00

03 Jan '00

> On Sun, Jan 02, 2000 at 12:31:00PM -0800, Owen DeLong wrote: > > Attempting to resolve a.b.c.d.e.f.g.h.i > > > > If you get back a return value from DNS that says > > "Authoritative answer, a.b.c.d.e.f.g.h.i does not exist" > > Bounce the mail. > > > > If you get back a return that says > > "Resolving a.b.c.d.e.f.g.h.i timed out without a return" > > Queue the mail. > > > > This also works for i, h.i, g.h.i, f.g.h.i, etc. > > I agree that this is the most sensible minimum functionality. What do you > break if you treat what you consider persistent terminal conditions as > transient? > Usually, if I send something to foo(a)delnog.org instead of foo(a)delong.org, I get a bounce back before I depart my MUA. You would break that. > Perhaps your business policy is to deliver all mail that can possibly be > delivered, making the assumption that everything's transient unless the system > has what it considers reasonable proof that it is indeed permanent (after > for instance several attempts spread out over a period of time). > I think that an authoritative nameserver saying "I know for certain that domain xyz doesn't exist." is reasonable proof that an error requiring intervention exists. > Personally, I don't like systems which do not have backup paths to deal with > transient conditions. And we do know that even "authoritative does not exist" > can be a transient error condition caused by a registry glitch etc. Sure, we > can all just ignore reality and assume we live in a perfect world, but.... > But a registry glitch is an error requiring intervention, therefore, I would much rather receive my mail back immediately so that I can intervene, than have it sit there while everything else breaks hoping that it will eventually clear itself. Owen > Cheers, > Chris > > -- > Christian Kuhtz Architecture, BellSouth.net > <ck(a)arch.bellsouth.net> -wk, <ck(a)gnu.org> -hm Atlanta, GA > "Speaking for myself only." >

1 0

got any good dirt on 12.0.X(S)?
by scott w 03 Jan '00

03 Jan '00

I'm looking for anything on IOS 12.0.X(S). The Good, the Bad, *and* the Ugly. Mostly the Bad and Ugly.... I haven't found too much in the archives. If I missed it let me know, I've got my flameproof panties on... 8-) Most likely, private replies are best. I'll summarize later if there is sufficient interest. Thanks, scott

1 0

BGP session resets
by Sean Donelan 02 Jan '00

02 Jan '00

In the last 48 hours I have received several reports about different providers resetting or reloading ALL of their external BGP sessions due to problems. I have not heard why the ISP reset their sessions, or details of what the problem is/was. At first I just assumed it was normal maintenance activity. ISPs do reload their BGP sessions on a regular basis. But as additional reports are coming in, it doesn't seem to be normal maintenance. The resets are ISP network-wide, and happening outside of typical maintenance windows. I have reports four different ISPs needed to reset their BGP. A small number, and it may just be because people are watching things a bit closer than normal. Does anyone know why ISPs are doing this?

1 0

Re: mail does bounce (was: Customers down?)
by Sean Donelan 02 Jan '00

02 Jan '00

On Sun, 02 January 2000, Owen DeLong wrote: > I'll take a stab at this: > > Attempting to resolve a.b.c.d.e.f.g.h.i > > If you get back a return value from DNS that says > "Authoritative answer, a.b.c.d.e.f.g.h.i does not exist" > Bounce the mail. > > If you get back a return that says > "Resolving a.b.c.d.e.f.g.h.i timed out without a return" > Queue the mail. > > This also works for i, h.i, g.h.i, f.g.h.i, etc. I'm happy to agree this is a very sensible thing to do. However, there is no requirement for any mailer to do it. Some mailers may, but other mailers may not. So it is not very wise to assume all mailers work the same as yours. There is a written recommendation that domains should have DNS servers on seperate networks. There is no such recommendation that mailers retry or queue mail when they can't reach DNS servers.

1 0

Re: mail does bounce (was: Customers down?)
by owen＠dixon.delong.sj.ca.us 02 Jan '00

02 Jan '00

> On Sat, 01 January 2000, "Steven J. Sobol" wrote: > > What I was referring to was a situation where the domain does exist in the > > root nameservers but the authoritative nameservers are unavailable. > > So .COM exists in the root servers? How do you decide to treat second > level domains (i.e. donelan.com) versus third level domains > (i.e. www.donelan.com) versus third level domains under country code > domains. > > > > > > I'll take a stab at this: Attempting to resolve a.b.c.d.e.f.g.h.i If you get back a return value from DNS that says "Authoritative answer, a.b.c.d.e.f.g.h.i does not exist" Bounce the mail. If you get back a return that says "Resolving a.b.c.d.e.f.g.h.i timed out without a return" Queue the mail. This also works for i, h.i, g.h.i, f.g.h.i, etc. Owen

3 2

an anticipated y2k situation
by scott w 02 Jan '00

02 Jan '00

My prediction was the NANOG mailing list would see a large increase in traffic due to y2k. It did. And here's one more... ;-) Happy new year/century/millennium everyone!! scott

1 0

RE: mail does bounce (was: Customers down?)
by owen＠dixon.delong.sj.ca.us 02 Jan '00

02 Jan '00

> > > > I think I like it better that way. Just because both nameservers are > > > temporarily down doesn't mean the domain doesn't exist. :P > > > > Seems to me you're talking past one another. If all nameservers for a > > domain are down there *is* no nameserver which can say that the domain > > is authoritatively non-existent. (OK, you could get a negative caching > > answer from one of the authoritative servers on the level above, but > > that's a different issue...) > > > > I agree with Greg Woods - if a domain is authoritatively non-existent, > > I'd expect a sane mailer to bounce the message. > > I'm glad that you don't write code, especially sendmail code. > > I'm hoping that some clarification of terms will kill this thread... There is a difference between the following terms: Authoritatively Non-existant domain: A domain for which the servers authoritative for the parent zone report that the domain does not exist. Unreachable Authoritative Servers: A domain which exists and posesses a delegation in it's parent servers, but which authoritative servers are currently not reachable. The former is regarded as a permanent error and mail is immediately bounced by every mailer I know of. The latter is a transient error and is treated as such by sendmail. I would hope it would be treated as such by other mailers. Owen

1 0

Re: [shadyron@total-bollo.cx: What happened in the UK]
by michael.dillon＠gtsip.net 02 Jan '00

02 Jan '00

On Sat, 01 January 2000, Joe Abley wrote: > We rolled over 1h35m ago. Rollover saw (locally) two of the four mobile > networks drop the users in the room with me (ie they saw coverage totally > drop). However, the other two networks: Orange & One2One were fine, with > users being able to make and receive calls fine. Attempting to call from > Orange -> {Cellnet | Vodafone} received either line busy or network busy. I was on the Westminster bridge near Big Ben at midnight and it was amazing how many people were pulling out cellphones just before midnight to phone up friends in other places like Sydney or New York. When people commented on being unable to make cellphone calls I asked what network they were on and most were on Vodafone but some were on One2One. Then about ten minutes into the fireworks display I decided to try making a phone call but my Orange mobile no longer had service so they were down too. I suspect this may be a combination of too many people in one small area (3 million in the streets of Central London) and too many simultaneous attempts at making a cellphone call. I think that cellphone use is a lot more common here in Europe than in North America. I sat on a train watching ten year-olds sending SMS messages to each other on their cellphones. As more and more such wireless devices are deployed, the possibility of localized overload becomes much greater. I experienced a similar overload four years ago at a technology conference held at a ski resort on top of a mountain which attracted more cellphone users than the single cell on the mountain could handle. Is this something that wireless TCP/IP can alleviate or is bigger infrastructure the only cure? --- Michael Dillon Phone: +44 (20) 7769 8489 Mobile: +44 (79) 7099 2658 Director of Product Engineering, GTS IP Services 151 Shaftesbury Ave. London WC2H 8AL UK

1 0