I noticed that I already started out-typing the system at 10 ms, however the supposed 60 ms keyboard latency of a Logitech k360 has never bothered me before.
That 10ms and that 60ms are pretty different things. I tried the 5ms setting on my Mid-2014 Macbook Pro and (measured with a high speed camera, plugged in with 50% battery, two browsers open with about 100 tabs between them but nothing consuming more than 10% CPU on the machine), the latency was 130ms - 170ms across a few runs, I’d probably see more variance if I had an automated setup and could easily get tail latency. I tried 100ms and got about 200ms total latency (only two runs at 100ms, it’s 2am local time and I should get to bed, I’d probably see a similar level of variance as with the 5ms setting if I tried more runs).
Jamie Brandon has a post on what happens when you try to specify precise delays in the browser. This isn’t the same thing since his post discussed delays when the input pipeline is effectively removed, but even without the input latency, a 10ms specified delay in the browser can easily cause 3x to 7x the actual delay, depending on what you’re trying to measure. As with the measurement I just did, it’s only a few runs, tail latency should be higher. If you want to “port” Jamie’s numbers to what we’re looking at here, Jamie’s delay is added to the input latency that’s already there with any off-the-shelf computer hardware.
Thanks for confirming my suspicions.
i felt right at home up to 200ms - standard experience of logging in to a terminal on a US machine from my crappy Greek internet + wifi.
You should try out mosh, makes ssh over crappy internet way better.
I once worked with a customer from Australia. That was worse than 200ms I think, I really feel bad for people inside Australia if it’s like that for them all the time.
(Australian here) Thankfully, I rarely have cause to SSH into a box that isn’t based in Australia. The few times I do, it’s pretty bad, but I’m generally only there for a minute.
I think years of using remote terminal connections on various flaky networks have hardened me to input delay. None of these annoyed me. 100ms made me nostalgic for the time when I would write school reports on Word in windows 3.1 on an 286.
I suspect these don’t bother me because I’m already on a 30ms keyboard and I don’t have the fastest machine, plus I’m used to typing on a machine under heavy load or over a high latency network connection. I’m just used to not having the totally synced feedback.
I thought it would all be worse, but it really doesn’t matter - I don’t need to look at what I’m typing as I type it, only after it’s been written, as to proofread.
I think it makes a massive difference whether you’re typing lots of text (and don’t really need the feedback from the screen), or editing, where you need to move the cursor about and wait for the feedback that the cursor is in the correct position (There may be exceptions, e.g. if you’re familiar with an editor like vim, you can type a sequence of commands with the confidence that it will move to the desired position and perform the required edit. However, a single mistake can completely alter the whole sequence of commands).
Even when editing I know what I should be typing before I type it; else it’s extra work for no reason.
30ms was OK for me despite being noticeable.