Hacker News Reader

Comment by Abishek_Muthian | Hacker News Reader

A moment of sympathy for the engineers who have to explain their non technical project manager on why their application is getting this error.

View on HN

I learned my lesson not put cutesey messages in the logs the hard way. I was young and dumb and in my case it was for a restore from backup system. A customer had failed hardware and was trying to restore from backup. It was failing and the funny message all of the sudden didn't sound too funny when customer was reading it back to me in their own voice over the phone.

Oh I’ve done that too as a rookie. Software displayed an error message when the application was in a state that was impossible for it to be in, at least in theory. The error message was a line from 2001 Space Odyssey: “I’m sorry Dave, I’m afraid I can’t let you do that”. Of course when it ended up getting displayed it was displayed to a customer called Dave who has never seen (or read) 2001 Space Odyssey.

The first lesson was to avoid funny error messages in the future. The other was that “impossible” things tend to happen surprisingly frequently in software.

With 7 billion people on the planet, one in a million events are guaranteed to happen to about 7k people.

Law of stats of large numbers etc

More like birthday theorem applies here .

There's a lot of different possible 1 in a million events, though...

That is an excellent point actually...

True. I really intended it as 7k people will have had the same 1 in a million event. Sorry that wasn't clearer.

"Million-to-one chances...crop up nine times out of ten." https://wiki.lspace.org/mediawiki/Million-to-one_chance

A formulation of this I heard recently:

As a lottery ticket buyer, you may not need to prepare for winning the lottery. But as a lottery commission employee, you had better prepare for someone winning.

> The other was that “impossible” things tend to happen surprisingly frequently in software.

"One in a million is next Tuesday"

https://blogs.msdn.microsoft.com/larryosterman/2004/03/30/on...

That article contains a link to (http://www.jumbojoke.com/000036.html), but the link is broken. I found it at (https://web.archive.org/web/20050205043510/http://www.jumboj...).

Wow, that's hilarious. Did he get creeped out at seeing his own name?

Yeah. Save the cheeky errors for internal tests.

Please don't even use it for internal tests... https://thedailywtf.com/articles/We-Burned-the-Poop

I meant internal test suite code. As in, the client / customer would have to be able to run my test suite (and likely watch it fail) to see the error code.

I've used "Nothing happens" (from ADVENT) when an unknown value was given for an "action" parameter. Appropriate, I still think, but a bit cute.

Yeah, good logging and good error reporting are those things to learn by being burned by a funny message at 3 am, or being embarassed by a funny message. Usually both at the same time.

It's amazingly frustrating to see: Hey if you just barfed the output of that failed command into a shell I could fix it but now I have to figure out why this several-hour command failed... bleh.

Not just logs. Recent Opera versions have taken to replacing error messages in browser, with cutesy animated spaceships and whatsnot. A error message i can live with, an uninformative cartoon just drives me furious while trying to figure out what broke.

Unless they're paying for NPM, my sympathy is limited. They should _really_ be putting Artifactory (or equivalent) between themselves and the public NPM registry. Or, at the very least, fallback package.jsons pointing to the source Github repos of each dependency. Anything at all so that npm install doesn't have a single point of failure.

I'm as technical as anybody who might be explaining this to anybody else, and I still don't understand why they're getting this error even after reading the linked thread and the comments here.

Reply on news.ycombinator.com