January 2000 - NANOG - lists.nanog.org

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

TDL .it maybe down???
by Giorgio Lembo 02 Jan '00

02 Jan '00

Hello, Has anyone noticed any strange behaviour in dns.nic.it? Thanks Cheers, Giorgio --- [root 17:15:02:giorgio]# nslookup Default Server: localhost Address: 127.0.0.1 > server 194.185.132.253 Default Server: venere-eth1.inet.it Address: 194.185.132.253 > www.iol.it Server: venere-eth1.inet.it Address: 194.185.132.253 Non-authoritative answer: Name: www.iol.it Address: 195.210.91.83 > www.nic.it Server: venere-eth1.inet.it Address: 194.185.132.253 *** venere-eth1.inet.it can't find www.nic.it: Non-existent host/domain --- Giorgio Lembo INS Ltd - Milano

1 0

ISP Y2K mailing list summary
by Sean Donelan 02 Jan '00

02 Jan '00

This was the final summary post for the NANOG ISP Y2K mailing list. Special thanks to Merit for allowing the ad-hoc group of Internet Service Providers to share the name for the NANOG ISP Y2K event. This post is a couple of hours early, because I want to go to sleep. NANOG ISP Y2K Summary Status 2-JAN-2000 12:30 UTC Y2K Summary Although the "hotlines" got a lot of press, much of the information came via e-mail. It was difficult to dial into the international conference bridges near midnight in various countries due to congestion on the international voice circuits. The Internet is a dynamic place. All numbers may vary depending on where you are on the Internet. But the trends should be the same. The Internet global routing table showed a slow decline in networks from 71,200 on December 25, 1999 to 70,911 on December 30, 1999. On December 31, 1999 the networks steadily declined to 69,812 by 1 hour before Midnight EST. During this hour the routing table quickly dropped to 67,795 routes, and held steady from Midnight until 5am EST. After 5am EST January 1, 2000 the number of networks started increasing steadily to 70,455 on January 2, 2000. The number of unique ASNs in the global route table declined from 6,380 on December 30, 1999 to 6,330 at January 1, 2000 Midnight EST. The number of terminating ASNs showed a similar decline from 5,180 to 5,130. Which leads to the conclusion transit networks did not disconnect during the rollover. Peak traffic levels across MAE-East were lower on Friday December 31, about 1.6Gb versus 1.9Gb for a normal workday. There was a slight but noticable 0.2Gb dip in traffic at Midnight EST. Followed by a slight increase after Midnight EST. All well below peak traffic capacity. Traffic flow followed the normal sine wave pattern, except for the dip at Midnight EST. Summary of ISPs 2 ISPs reported a routing problem, isolated to normal circuit problems. 1 ISP reported voice-call (POTS) congestion in New Zealand shortly after midnight through several voice carriers. Internet backbone connectivity was not affected. This was the pattern through most time zones. Several ISPs reported trouble synchronizing NTP servers. Congestion on the NIST network is believed the cause. NTP servers elsewhere did not show any problems. 1 ISP reported messages sent in elm showed year as 100 instead of 2000. Patch was available prior to Y2K. Several reports of user-written scripts containing poor date handling practices. Report of one web site defacement after close of business. Report of some unauthorized domain name transfers at the beginning of the holiday weekend. Not unusual, seems to happen every holiday weekend. Two ccTLDs were down for part of the rollover. 13 ccTLDs had partial problems. No Y2K faults were found. Circuit problems, administrators turning off servers, and ordinary DNS problems. Exchange points worldwide reported a decrease in traffic immediately before their midnight local time. A sharp increase immediately after midnight local time, but less than peak capacity. Certain "event" web sites saw traffic increase 20 times over normal levels. Some event sites report being overwhelmed. But overall, traffic appears to have been lower than peak business day capacity. The use of many country-specific information sites, and the rolling nature of the Y2K event helped distribute the traffic worldwide. In general public interest waned about 30 minutes after local Midnight in each timezone, leaving more bandwidth for each successive timezone. Midnight US Eastern time did dominate changes in traffic and routing. A few rumors sprung up throughout the night. Most were quickly determined to be unfounded. But overall, not as many as I thought would spread. Issues known/predicted prior to the Y2K rollover Some sites plan to shutdown or disconnect from the Internet over the New Year's weekend Some certificate authority certificates expire on December 31, 1999 Increase in voice and cellular calls immediately around local midnight may cause some congestion on circuits Information from the media (Note: ISP names are from publicly announced information. Names will not be included from non-public sources.) C I Host (www.cihost.com) issued a press release about a name server problem. No client data files were damaged. The data corruption that occurred Dec. 29 is isolated to the [company local] nameservers only and the restoration remains unrelated to any Y2K issues. Keynote Internet Performance Update #2 reports slower access to New Zealand after rollover, 6.2 seconds. No information what the measurement was before the rollover. Appears to be localized to specific web sites (i.e. congestion) France's National Weather Service Internet site had a display problem with a date. The page shows 01/01/19100 instead of 01/01/2000. United States Naval Observatory web time site had a display problem with a date. The page shows 19100 instead of 2000. Hacker target Japan Y2K Center. No intrusion or damage reported. Several stories based on Keynote press releases report the Internet passed through the Y2K rollover unaffected, other than a few spots of congestion around particular web sites. The www.y2k.gov web servers were predicted able to handle 40 million hits. Only 3 million showed on January 1. Many, MANY, ISPs, web hosting, and access providers issued press releases throughout January 1, 2000 announcing they had no Y2K problems (and I think they all sent me a copy). Summary from the Y2KCC/JP The number of reported troubles 25 Troubles with Stratum1 NTP server 4 Trouble with ICQ 1 Troubles with ccTLD 15 small partial troubles 13 others 2 Trouble with router 1 Trouble with Y2KCC/JP system 1 Latency of a particular site 1 Suspect of cracking 1 Trouble with NNTP server 1

1 0

UK GMT roll over
by Neil J. McRae 02 Jan '00

02 Jan '00

GMT and CET rolled over without any major incidents. The cellular networks were busy but thats to be expected. Regards, Neil. -- Neil J. McRae - Alive and Kicking. neil(a)DOMINO.ORG

12 13

Netscape JavaScript getYear() problem? Also bad JavaScript on CNN.
by Will Pierce 02 Jan '00

02 Jan '00

Hello folks, I discovered to my utter delight that my lights stayed on during the roll-over. However, after poking around looking for the year 1900, I only found one issue. It appears that the javascript function "getYear" in the Date class is inconsistent between Java implementations. This happens when getYear() is called with no arguments, in other words, when getYear() is queried for the current year. I am not terribly JavaScript "Clue-Compliant" so if I make any blunders in this post, please set me back on track. Specifically, I get the following test results: Platform Browser Version getYear() returns -------- -------- ----------- ----------------- Win98 Netscape 4.61 100 Win98 Netscape 4.7 100 Win98 MSIE 5.00.2919 2000 You can look at what your browser's JavaScript interpreter returns for getYear() by loading: http://www.dreamscape.com/willp/y2k.html Apparently, Netscape 4.61/4.7 is not properly reporting the year in a getYear() call. >From what I've read on the net (which is NOT extensive and I'm in no way shape or form a java guru)- the getYear() function should return a 2 digit date up to and including the year 1999. However, in the year 2000 and beyond, getYear() should return a 4-digit date. Now, whoever thought up this getYear() function behavior should be shot, IMNSHO. I saw this when my "CNN Ticker window" in MSIE5 indicated that the current year was 2300. The CNN JavaScript assumes the getYear() function returns the # of years since 1900. Caveat: I have not tried looking at the return values of getYear() in JavaScript on anything other than my own home computer. It is possible (though unlikely I hope) that what I've seen is due to a strange problem with -just- my computer and its mangled DLL files under Windows 98. I would appreciate any confirmation of this strange behavior of getYear() under Netscape and/or any help in understanding this problem. If someone could confirm that they too see the year 3900 on the CNN "ticker bar" and that these results -are- in fact problems, I would really appreciate the feedback! The risk: Any web sites that use the getYear() function will need to accomodate various return values of getYear() and properly distinguish the bizarre results that some browser versions will return. In the case of sites that only use getYear() to *display* the date for the edification of the user, this won't have any operational effects. However, for sites that use date processing in a mission critical situation, there may be some problems. While I have strong hopes that no one would use client-side JavaScript in a mission-critical application, there is still the possibility that this will be a problem on some websites. For most sites affected by this, it will only cause embarassment. Random List of Affected Sites (found so far): CNN http://www.cnn.com/ The CNN "ticker" toolbar (downloadable at: http://www.cnn.com/browsers/explorer.bar/ ) indicates to me that the date is 3900, because their javascript function "ShowDate" adds 1900 to the result of getYear(). (MSIE5.00.2919.6307 on Win98) The date is only used for display purposes (as far as I can tell). Computer Technology Associates http://www.ctai.com/ The date displayed on their homepage indicates "January 1, 192000" under MSIE 5 on Windows 98. Their JavaScript assumes that it is always the 1900's and that getYear() always returns the last two digits of the year. The date is only used for display purposes. Rocky Mountaineer Tours http://www.rkymtnrail.com/ Same problem as for CTAI above. The date is only used for display purposes. JJournal Jeep Owners Resource http://www.jjournal.net/ Same problem as CTAI and Rocky Mountaineer Tours. The date is only used for display purposes. PegasoWeb Web Site Promotion http://www.pegasoweb.com/ Again, year is shown as 192000. The date is used only for display purposes. Web Medicine http://www.webmed.com/ The "last date modified" on this site indicates that the site was last updated on January 1, 3900. Apparently the date is only used for indicating to the visitor that the website is modified daily. Sport/Utility Vehicle Owners Resources http://www.suv.com/ Displays current date as year 192000. Date only used for display purposes. World Trade Center Institute http://www.wtci.org/frmain.htm Displays current date as year 192000. Date only used for display purposes. Inland Empire Online http://www.inlandempireonline.com/business/index.shtml Displays current date as year 192000. Date only used for display purposes. Resources related to this: Overcoming Different Year Conventions - Doc JavaScript http://www.webreference.com/js/pharmacy/article1/overcome.html Y2K JavaScripts http://server9.hypermart.net/ucprogramming/javascript/article1.html JS Year 2000 Very Basic Test Page http://www.nikwilliams.co.uk/jsdate01.html Java City 2000 - Year 2000 Fix http://www.jc2k.com/y2k.html JSBible Calendar Scripts and Y2K http://www.dannyg.com/calendarY2K.html -- Will Pierce Systems Programmer Nuts & Bolts Interactive, Inc. willp(a)nbinteractive.com P.S. If someone could confirm this behavior as being due to 1. differences in the implementation of the getYear() javascript function, and 2. poorly coded JavaScripts, I would be very grateful for the confirmation. I am no JavaScript programmer by any means, so this whole post could be way off base. I am only posting it so that some more experienced Java guru can either correct my mistakes or confirm and expand on what I'm seeing. Flames to /dev/null pls.

3 3

Re: mail does bounce (was: Customers down?)
by Sean Donelan 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.

2 1

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

02 Jan '00

On Sat, 01 January 2000, Jim Duncan wrote: > I recommend you not use three-digit RFCs as authoritative references. > I agree with your line of reasoning, but current arguments should be > supported by current observations and decisions, as documented by current > standards. I see a lot of crap on the net that's crap fundamentally > because its development was based on a three-digit RFC. :-( Whether or not you use three-digit RFCs as authoritative references, other developers do, so you should expect to deal with such behaivor. Too many people think the motto be conservative in what you send and liberal in what you receive only applies to the other guy. He should never send something I don't expect, and should accept anything I decide to send. > The human element causes many problems, and yet it is the ultimate > solution for all of them. Rather than strive solely to make things > perfect, try to make things as close to perfect as possible *and* put some > thought into what you will do when things fail. Assume that "calling the > human element into play" is a part of making things perfect, whether > you're dealing with failure of timely delivery of a message to the > intended recipient or failure of network connectivity to fail over > gracefully to another path. The question is how long do you let things wallow in a state of computerized uncertainity before you call the human. Things fail, and sometimes people create situations which make the failures even worse. Do you immediately inform the human, try a few times and then inform the human, keep trying until the human notices its not working. Different people make different choices. If a mailer wants to queue mail on a DNS server failure, that is acceptable. But you shouldn't expect or depend on that behaivor. If you disconnect all your name servers from the network; some applications can, and therefor you should expect they will, treat it as a permanent error. Whether you think its the fault of the human who wrote the application, or the human who put all the name servers on the same subnet; it shows problems are rarely due to a single mistake. Rather a combination of decisions, which may have appeared justified alone, when taken together cause trouble. Its like the backhoe problem. There would be no affect on customers if either a) carriers provisioned sufficient alternate capacity OR b) a backhoe didn't cut the fiber. Its the combination of insufficient diverse capacity AND a backhoe attack which causes the problem for the customer. Unfortunately, the first instinct is to blame the other guy. So telco's blame the backhoes; and the excavators blame the telco's.

1 0

Traffic statistics from major ISPs
by Sean Donelan 02 Jan '00

02 Jan '00

In the voice world, AT&T, SBC, Bell Atlantic and others have released traffic figures about how many voice calls they handled around New Year's. I have not seen similar data from any major ISPs about traffic volume around New Year's. I know about third-party sites such as Keynote, and MIDS; but they don't measure traffic volume. I'm looking for basic data such as ANS used to publish for the NSFNET. The closest I've found is Abovenet's MRTG pages, but it is spread out over different links and its hard to aggregate the numbers without double counting, or undercounting. I know several major ISPs collect the data internally, but haven't released it like the voice side of the house.

2 1