Thank you for the fascinating writeup! How long did the whole thing take to put together?
From a reddit "expert", so I guess that answers that ;)
91–100 of 255 posts
Thank you for the fascinating writeup! How long did the whole thing take to put together?
From a reddit "expert", so I guess that answers that ;)
Why use Redis and multiple machines instead of keeping it in RAM on a single machine? I'm not claiming the Reddit people did anything wrong; they have a lot more experience than me here obviously. I'm just trying to figure out why they couldn't do something simpler. 333 updates/sec to a 500kB packed array, coupled with cooldown logic, should have a negligible performance cost and can easily be done on a single thread…
Because machines go down. If you don't expect your hardware to fail at the most inopportune time, you'll be screwed when (not if) it happens.
The front-end UX for scrolling that bitmap was quite frankly horribly badly designed.
What was wrong with it?
The front-end UX for scrolling that bitmap was quite frankly horribly badly designed.
What was wrong with it?
It's still there: https://www.reddit.com/r/place/
They have may have fixed one or two scrolling issues since, but the main issue is that if you
a) press LMB
b) move the mouse, and move outside of the pixelized area
c) release LMB
.. it does weird things.
Earlier quoted context omitted.
What was wrong with it?
It gets weird when the cursor leaves the box while dragging. Now, when you go back inside, you're still in drag mode since the box did not get the "mouse up" event and you end up selecting and dragging random text.
Myself, I just get so frustrated about the idiocy.
Why use Redis and multiple machines instead of keeping it in RAM on a single machine? I'm not claiming the Reddit people did anything wrong; they have a lot more experience than me here obviously. I'm just trying to figure out why they couldn't do something simpler. 333 updates/sec to a 500kB packed array, coupled with cooldown logic, should have a negligible performance cost and can easily be done on a single thread…
> 500kB packed array, coupled with cooldown logic, should have a negligible performance cost and can easily be done on a single thread. That thread could interact via atomic channels with the CDN
That's not simpler.
We used tools that we're already using heavily in production and are comfortable with.