Hacker News Reader

Comment by geerlingguy | Hacker News Reader

I lament the fact error messages are rarely, if ever, displayed anymore. Just a generic message in its place, usually not even indicating whether it's a 'you' problem or a 'them' problem :(

View on HN

It amazes me how most smartphone apps- which operate in a wireless environment of likely flaky connections, don't seem to be tested or designed at all to handle loss of connections. They tend to just become non-responsive or do erroneous things like play the first second of an audio file and then auto-pause, without reporting any errors.

Sooo like the Spotify app for the past 10 years.

These 1 percent of edge cases take so much effort to test and handle though.

1 percent is extremely frequent though. Like literally thousands of times per second on prod if it's a b2c application

A smartphone having a flakey internet connection isn't an edge case.

Exactly, it's a daily thing for users- it's only an edge case for the developers, because they are likely developing and testing with hard wired high speed connections.

Good chance most of the phones are virtualized as well to test across OS builds and that the actual deploy to real phones aren't tested nearly as well.

Personally, I despise cutesy error messages that try to make the program sound more human. "Oops, we're sorry this happened! Would you like a nice cup of hot cocoa while we try and figure out what happened? Or should I read you Goodnight Moon" again?"

Spotify is in offline mode...

This bug has been there since the first time I installed Spotify on a smartphone and it's still there a almost 15 years later.

Yeah, but at least it's a bit understandable that its changed in that direction, and usually you can get a developer-friendly error message if you look at the HTTP responses in the devtools.

I remember one time a company I worked at received a support message where the title was (paraphrased) "DONT HURT MY CHILDREN" from some person who saw an error message saying something about "couldn't dispose of child" or something similar, when the frontend broke. "Child/children" being kind of common in programming, I'm sure others faced similar scenarios.

I'm not sure if they were genuinely scared or just decided to have some fun with us while reporting the issue, but after that we made the error messages even more generic, so nothing could be misunderstood.

I guess it is this quest to not alienate the lowest common denominator that is the reason behind the stupidification of error messages. On one hand, people won't get scared, on the other hand, people get less context about why the error happened in the first place.

Wonder how many of the average users actually care?

The original Mac would pop up a dialog with a threatening icon of a bomb with a lit fuse, whenever it crashed! https://en.wikipedia.org/wiki/Bomb_(icon)

>The Bomb icon is a symbol designed by Susan Kare that was displayed inside the System Error alert box when the "classic" Macintosh operating system (pre-Mac OS X) had a crash which the system decided was unrecoverable. It was similar to a dialog box in Windows 9x that said "This program has performed an illegal operation and will be shut down." Since the classic Mac OS offered little memory protection, an application crash would often take down the entire system.

Unfortunately, the Mac's bomb dialog could cause naive users to jump up out of their seat and run away from the computer in terror, because they though it was going to explode!

https://www.youtube.com/watch?v=zQGX3J6DAGw

And Window's "This program has performed an illegal operation and will be shut down" error message was just as bad: it could cause naive users to fear they might get arrested for accidentally doing something illegal!

Generally error messages get watered down to nothingness because it gives news reporters less meat to chew on. Not knowing if its a "me or you" problem means that small problems might actually get missed and not need any PR response. Details won't leak and produce all kinds of speculation over the causes, etc.

You have to be careful with error messages. It's possible to give too much away and enable security vulnerabilities.

I'd assume the security vulnerability is there no matter what the error message says, but I guess very explicit and verbose error messages might expose details to make it easier to find said vulnerability.

There are so many microservices everywhere with different teams responsible, that if just one team doesn't forward errors properly it is messed up.

Integration tests are a thing.

You mean e2e from mobile app? Because they would have to be able to trigger those failure situations from all layers of microservices for which they have no control over.

Reply on news.ycombinator.com