The Y2K fail finally happened?

I got this 31 Dec 2020 at 1155PM

Try it again. It could be some scheduled downtime.

It figures, Amtrak 21 years late, a new record.

[OT] In early 1997, a mutual fund system (not the one that descended from the Kansas City Southern railroad) incorrectly withheld a million $ or three and forwarded it to the IRS. The system used two-digit years throughout, and they had recently hired an Indian software company to remediate their systems and had told their IT staff to NOT fix the problems themselves, leave it to the Indians who would fix it within six months. Unfortunately, exemption from withholding certificates are good for three years and to simplify, 00 - 97 < 00 therefore expired.

DEC (Digital Equipment Corporation) VMS operating system has and/or had a year 2038 "bug".

Hopefully, if Amtrak's system has failed, they will install something more user friendly. The current system is confusing and difficult to navigate.

Their site is back up and running ,may have been a year end backup was created at that time.

Backshophoss wrote: ↑Fri Jan 01, 2021 8:09 pm Their site is back up and running ,may have been a year end backup was created at that time.

That's my guess as well. It probably started some pre-scheduled daily maintenance at that time and the timing was perfect to get the error.

The Y2K problem was and still is very real. Depending on how the programs were revised, it may happen again in 2028!

In the late '90s, I was a contractor at a Fortune 500 firm engaged to upgrade/replace an outdated predecessor of IBM CICS online programs with the CICS equivalents. I wrote a translator program on my home PC to read the COBOL programs, convert all INTERCOMM aspects to CICS, and output a 'new' program. One of the groups of programs I converted had a 'future revenue' calculation that incremented the current year by 1 for the next 3 years. Needless to say, when it rolled over from '99' to '00' the calculations went wild and caused the program to loop until cancelled.

Luckily, or unluckily, that company had employed a small 'army' of contractors to handle all Y2K issues starting about 1997. They used a small sub-routine program to add 28 to every year resulting in dates 01/01/2000 through 12/31/2027 and 'return' those dates which were used for number of days between calculations. I can't recall the reason why, but as of 2028, the Y2K problem will return with these programs if they haven't been modified with all date fields in data files and data bases expanded to a 4 digit year. I simply incorporated 'their' sub-routine into the programs I was converting thereby accomplishing Y2K compatibility at the same time.

Why 28 you might ask? That's because the calendar exactly repeats every 28 years, including leap years.

Last time I checked with a couple friends still at that company about 15 years after my project was complete, their master data files and data bases had been revised to 4 digit years. I presume that the sub-routine program is no longer used. However, given that after a corporate merger that happened while I was there, all together there's about 3500-4000 programs involved. Hopefully, some of those have been eliminated, but there still has to be a 'mountain' of programs to go through and fix.

There may have been something more widespread going on. Shortly before midnight on New Year's Eve I was looking in our bank account for our monthly retirement annuity payment, which usually arrives on the last day of the month. I got a message that there was some trouble with online banking. I mentioned it to somebody on Jan. 2, and he said that he too had trouble with his credit union online on Dec. 31.

Literalman wrote: ↑Sun Jan 03, 2021 5:02 pm There may have been something more widespread going on. Shortly before midnight on New Year's Eve I was looking in our bank account for our monthly retirement annuity payment, which usually arrives on the last day of the month. I got a message that there was some trouble with online banking. I mentioned it to somebody on Jan. 2, and he said that he too had trouble with his credit union online on Dec. 31.

It may not be related, but I can lend a bit more proof to the funkyness. NYE I get paid, but I'm already slightly in the negative. Available positive, and I can transfer against that and pay bills and such. Balance goes more negative, available is positive, paycheck clears New Years... and my bank pre-zeroes out the negative balance with a "Temporary Overdraft coverage" deposit. WTF? I've never seen that!

[OT] What is INTERCOMM? These problems with Amtrak's website and banking websites at the same time may be coincidence, or they may be using the same cloud vendor, etc. Stuff happens at year end that may not happen other times of the year too. As for Y2K, the companies did not take it seriously until they required the corporate officers to swear under penalty of perjury that they had resolved any Y2K problem. Then suddenly it was job #1

Literalman wrote: ↑Sun Jan 03, 2021 5:02 pm There may have been something more widespread going on. Shortly before midnight on New Year's Eve I was looking in our bank account for our monthly retirement annuity payment, which usually arrives on the last day of the month..

Mr Literail, my RRTA Annuity hit my bank Jan 2; and within twenty cents of what I had calculated the new amount to be.

Still have not received my 2021 "love letter" from the RRB detailing new amounts for Tier I, II, Med B, and IRMA; since my "industry after life" was that of a CPA, that precise calculation from the RRB is of interest. (addendum: got it in mail today)

Incidentally, there is a topic regarding such over at the Employment Forum.

dgvrengineer wrote: ↑Fri Jan 01, 2021 6:58 pm Hopefully, if Amtrak's system has failed, they will install something more user friendly. The current system is confusing and difficult to navigate.

Seconded. For instance, you really have to have some knowledge of where your closest Amtrak train stations are to your origin and destination. It would be nice if you could input your origin and destination, which don't necessarily have to be Amtrak-served locations. For instance, you input Town B, and a message pops up saying "Sorry, we don't serve this town, but here are your closest Amtrak stations: Town A (ABC) is 10 miles from your location, Town C is 30 miles from your location". For instance, you type in "Bennington, VT to Montreal, QC" and the message says "Sorry, we don't serve Bennington but Albany is ## miles away and Saratoga Springs is ## miles away... would you like to choose one of those?"

Maybe the technology to make that happen didn't exist back in the early web days, but it should today, I would think!

BandA wrote: ↑Sun Jan 03, 2021 11:57 pm [OT] What is INTERCOMM?

Intercomm predated IBM's CICS by several years. I think it was created by a company whose name I can't remember about 1970, plus or minus a couple years. By the time the conversion project started in late 1994 with a pilot conversion, I was told there were less than 10 Intercomm users left, one of them being Social Security. Having dealt with federal software contracts in prior years, I chose not to give them a call at the end of the conversion. I made enough money on that conversion project to retire the first time at age 51.

Although there were separate MMU screen definitions similar to BMS in function, MMUs contained every 3270 'internal' component such as length, attribute and MDT positions. Intercomm programs had to send & receive the entire 3270 buffer every time. In COBOL, that required lengthy 01-level definitions that were populated as desired then sent out. The real 'trick' in the conversion was to map the BMS-generated names to the data field names used by Intercomm and move all the data to/from each as needed.

For what it's worth, I still consider the MMU to BMS conversion subroutines in the translator the most difficult and complex code I ever wrote...in COBOL on my PC, no less!

bratkinson wrote: ↑Fri Jan 15, 2021 10:02 pm Although there were separate MMU screen definitions similar to BMS in function, MMUs contained every 3270 'internal' component such as length, attribute and MDT positions. Intercomm programs had to send & receive the entire 3270 buffer every time. In COBOL, that required lengthy 01-level definitions that were populated as desired then sent out. The real 'trick' in the conversion was to map the BMS-generated names to the data field names used by Intercomm and move all the data to/from each as needed.

Mmmm 3270s. That brings me back to the college days, when I was working on campus with the in-lab IT support teams, back in 1995. Email for the team was on a "big iron" VM370, while students got normal Unix accounts and had to deal with "pine" on DEC Alpha hardware or even NeXTstep computers. Slowly they switched over to Solaris, and I think they're switching to Linux now.

Back to Amtrak, though. Lets see who the provider actually is... actually, we can't (at least from probing DNS and IP addresses). Amtrak uses an reverse proxy provider called Akamai. This acts as a bit of a firewall to filter out some (a lot) of the crap web sites get hit with that isn't legitimate people. The error that John got was from Amtrak's servers behind the Akamai firewall, not from Akamai itself (which would have different site branding).

That said (putting on day-job hat for a bit) I won't be surprised if Amtrak has it's own data center, with a database server cluster and a virtual machine host cluster for the web work, maybe with a "middle API" layer or two for security separation (PCI compliance, probably).