Comment by ndriscoll

If it's only a few minutes, you could just as well fix it at the start of the day. And if it's not, oh well? People can still read books. The librarian can still help them to find a book. If they want to manually track until the computer is back up, they could even still do checkouts.

If those users just say "oh, okay, I'll check back in a couple hours", then okay. Again, this is not hundreds of thousands of people without power or water or something. It's viewing or uploading to an archive. Like, if you're a user, do you really want someone woken up at 3am because you can't use the wayback machine right now? Because "they should take it seriously"? There's a person actually having to deal with that because... you really want to get that one Grateful Dead concert video from their free archive right now? I'm happy to say that yeah I hope their engineers aren't on-call after hours. I would much prefer that they get a good night's rest rather than fix it for me. I hope they're a cool place to work without delusional, psychopathic management.

Perhaps put another way, is there any computer service now that doesn't need a guarantee? Or is everything just critical now?

My point isn't that on-call is overcomplicating, it's that inventing and chasing an SLA often leads to expensive complicated systems (which, ironically, can often directly lead to lower uptime), and on-call has a human cost.