Broken Netcode Demonstration -- A Bogus Kill

by **Jediluke** » Mon Oct 28, 2013 10:50 am

I'll help you test.

And I'm 40 up and 40 down...connect should be consistent.

by **Drakona** » Mon Oct 28, 2013 11:49 am

Flare fight with Jediluke over a 50 ms connection. Retrohomers + long packets.

Pretty much identical performance to LAN, as expected. Definitely an improvement, but still not as good as 0.55.

I'll continue to see what I can learn from the code.

by **Ryguy** » Mon Oct 28, 2013 12:44 pm

More amazing work.

by **DKH** » Mon Oct 28, 2013 12:53 pm

I'm impressed Drakona, nice job. When I first came back to Descent in 2009 (Before Jeds) we were all using .53 rebirth. To this day.. Diablo swears by it and has always said the newer versions like .57 .58 are messed up. He also has enough experience to see what Jeds sees. I can also tell a difference. Anyways, I just wanted to give you some support! Nice work!!! --DKH

by **Ryguy** » Mon Oct 28, 2013 4:16 pm

I'm curious about the long packets in 0.58.1. At one point, someone had said something about it sending more rotational position information. I'd be curious to see if it decreases the problem.

by **Drakona** » Mon Oct 28, 2013 5:04 pm

Good thought!

Here's 58.1 with short packets enabled:

. . . looks a lot like 57.3, though not quite as pathological as the result Ryusei and I got.

Here it is with the short packets diabled:

. . . MUCH better! Though there are still some puzzling exceptions off by three degrees or more. So there's still something else going on compared to the rock-solid 0.55 performance.

(Both those tests were vs. Lotharbot on LAN, but that's not seeming to make a difference).

Based on this data, it looks like 58.1 with short packets disabled is our best version for avoiding rage-quit-inducing losses. Though it's kind of a pick your poison thing -- primaries not working or extra evil homers? People keep pestering me to do a retrohomers mod of 58.1, and I keep saying we should just wait until 58.2 comes out. However, this data does make that look like a worthwhile effort. Plus, if I put in the time, by murphy's law, that will cause 58.2 to come out, which would be a good thing.

I'll keep digging. Those exceptional shots could certainly cost you a match if they happen to be smart missiles.

by **Ryguy** » Mon Oct 28, 2013 8:52 pm

Thanks for looking. Interesting results. I wonder how well a large game would work with short packets disabled.

by **Drakona** » Mon Oct 28, 2013 9:01 pm

Gave it a hard test with a DCL match vs Lotharbot in Athena GV no missiles 58.1 / long packets. First time I've played him that the connection felt truly lanlike. Big improvement!

by **JinX** » Mon Oct 28, 2013 11:27 pm

Can I get a tl;dr in a new post that covers everything discovered thus far? That'd be awesome. =)

by **Drakona** » Tue Oct 29, 2013 9:14 am

So it came to pass that Drakona discovered that short packets were evil, and the source of much suffering and woe. She did declare that the versions of the game which only offered short packets were terrible and unfit for competitive play. Thus did she recommend 58.1 with short packets disabled, all the while weeping, for there was no version of the game which contained both correct homers and reasonable orientation data.

And some mystery still remained, as even 58.1 did not match the performance of the legendary 0.55 version, and Drakona did curse the Rebirth gods and vowed to destroy what they had wrought.