3 ms·
I don't think you understand what I wrote. Let me try to rephrase. All of these protocols are tightly bounded in size. There is no way for them to get to 120
by felixgallo 12y ago
I don't think you understand what I wrote. Let me try to rephrase. All of these protocols are tightly bounded in size. There is no way for them to get to 120D over time. What we're talking about is, under packet loss conditions, moving from 20 (IP header) + 8 (UDP header) + 1 byte of command data + N bytes of state/overhead, to 20 + 8 + X bytes of command data + N bytes of state/overhead. N depends on the application, but for my implementation, generally it's around 20. So we're going from 49 bytes to, usually, around 50 bytes, and the absolute max for X is, say, 32; which would be 48 + 32 = 80 bytes, or a little less than 2D.
Now, in a fictional universe where a game protocol operating at 20 updates per second and putting down 50 bytes is causing network congestion, I happen to agree with you. That would be terrible! Never, ever, ever happens, ever, not even .0001% of the time, ever; not ever. And further, never will. But I've only been doing this for 30 years.