I recently went through a final-round interview process, and a large piece of the technical part of the interview involved handing me a laptop connected to the internet and the interviewer saying "Build a REST API that can handle these 4 curl requests and respond correctly. Use any technology you want. You have access to Google and anything else you can find. You have about an hour, I'll be back to check on you in 20 minutes. Go!"
I felt like this was SUCH a better test of my abilities as a professional programmer versus remembering specific algorithms or reciting "best practices" for X, Y and Z.
Oh man, I'd probably spend the first half of that time adjusting to somebody else's development environment, and the second half troubleshooting all the runtime errors caused by various ':w's and 'u's scattered throughout the code (unless it were an OS I was familiar with, I'd hesitate to implement this in a language that requires configuring a compiler & build system. Fortunately, most web software seems to already be written in portable languages).
The better solution I've seen to this problem is for them to warn in advance that something like that will happen, suggest bringing your own machine with you, and point you at a skeleton git repository to start working on.
I've made this comment before, but people don't seem to understand how used one can get to a specific setup.
I've spent for example 20 years working with MY Visual Studio setup, with a normal desktop keyboard and mouse. Then I get handed that crappy flaptop that no one wants to use.
You do what you can, but still your effectiveness basically drops by 40%.
This was a great test - I just gave myself this test (I'm usually working in embedded systems so this wasn't something I do all the time) - completed in 30 mins, most of it waiting for installs or reading the web. I think it might be possible to add postgres and OAUTH2 into this as well (assuming one can grok OAUTH2 in 30 mins...).
I felt like this was SUCH a better test of my abilities as a professional programmer versus remembering specific algorithms or reciting "best practices" for X, Y and Z.