Broken Netcode Demonstration -- A Bogus Kill

by **Drakona** » Tue Oct 29, 2013 5:45 pm

And I'm also thinking if I have programming time to burn, I should focus on figuring out why the tracker isn't working in 58.1. Now that I'm much more interested in playing that version, I suddenly notice how painful it is to write down IPs. The tracker seems to be seeing the games, but Descent seems to be having trouble reading the game list from the tracker. Surely that's fixable.

by **RiTides** » Tue Oct 29, 2013 9:36 pm

That would be awesome, as yeah I find writing down the IPs pretty painful, too!

by **Drakona** » Wed Oct 30, 2013 9:06 am

A day of playtesting later, I'm hearing that 58.1 retro long packets is definitely better than 57.3, and certainly the best thing we've got . . . but it clearly isn't perfect. Perhaps once or twice a match, you'll see something clearly bogus. Which is certainly better than bogus being normal, but I'd like to do better.

by **Drakona** » Wed Oct 30, 2013 11:20 am

I've created a mod I'm calling 'sniperpackets' -- it sends a high fidelity forward vector in the shot packet. Roughly doubles the size of the shot packet.

This is not original Descent. It's something I cooked up . . . but the aim is to make it better than the original. It's like how Rebirth takes advantage of modern hardware to increase the screen resolution of the game. Sniper packets takes advantage of modern fast internet connections to give more accurate gameplay. (Short packets was perversely the opposite philosophy . . . the equivalent of making the screen resolution SMALLER on modern hardware.)

Flare fight test results: (58.1 long packets retro homers sniper packets on LAN)

Those are perfect zeroes. I was thinking the demo recording apparatus might introduce some errors, but nope -- the shot orientation was IDENTICAL on both screens, down to the limit of what a demo records. By design, it should actually be identical down to the bit inside the physics engine.

Will there be drawbacks to running such a build? I don't know. I don't seriously expect doubling the size of the shot packets to cause network problems--it just isn't that much data. Even with 8 players firing vulcan, this is a data increase of 1.9 kBps. For the host, 15 kBps on the upstream. I just can't imagine that that's so much as a speed bump in 2013. But I won't know for sure whether f I broke anything else in the process until I get a test game in. (Firing 100 flares, walking across the room to my husband's computer and doing it again . . . is not a test game.

)

I am concerned that since the original game was never this accurate -- not even over LAN -- this might overpower highly accurate guns like gauss and vulcan. We will have to see.

However, I do know that some high level players have setups that intentionally sacrifice turning speed for extreme shot accuracy -- it would be a shame if the net code didn't give you the benefit of that accuracy.

I'll be looking for help testing later today, but I'm hopeful.

by **Jediluke** » Wed Oct 30, 2013 1:10 pm

I'll help test in about an hour or so.

by **RiTides** » Wed Oct 30, 2013 1:21 pm

Wow, that looks promising. And the name "sniper packets" is just great.

I am hoping to be able to get on some this weekend and I'd love to try out this version with you then :)

by **Drakona** » Wed Oct 30, 2013 4:37 pm

Sniper packets looks good. I'm ready to release it.

D1:
D2:

D1 source:
D2 source:

I had help testing from Jediluke, Ryusei, Bahamut, Icewolf, Swoop, and Mark392. We gave it some pretty tough testing in 6 player anarchy in both D1 and D2; no noticable network effects. Icewolf and Swoop helped us look at it under laggy conditions. Ryusei confirmed that it compiles and runs on Linux. Jedi and Baha gave us some expert pilot feedback. (Mark wasn't in game for very long).

The major feedback is that the shots are amazingly accurate with the sniper packets. Vulcan is much more dangerous than it used to be, and -- according to Bahamut -- so is fusion.

Definitely better than 58.1 even with the long packets, but something so accurate is definitely different than what you're used to. I recommend checking it out in for-fun games first!

------------

I have not yet figured out what's going wrong with the tracker. When I do, I'll offer an update, but it could be a few days. I made a stab at it yesterday and didn't figure out what the problem was.

by **RiTides** » Wed Oct 30, 2013 4:57 pm

That's awesome and amazingly fast work, and a nice test group to boot. Looking forward to playing it!

by **ryusei117** » Wed Oct 30, 2013 7:24 pm

I just scrutinized the Sniperpackets demos I took today. If the ships are firing shots at different angles than they're facing, the effect is so slight I can't even see it in a frame-by-frame playback. Even Icewolf, whose total ping to me was around 300 ms (him to host to me), was shooting straight on my screen.

I feel like the holidays have arrived two months early.

As for the tracker issue, I think Icewolf might know what's wrong, or at least have some insight on the matter. He posted some information , and Zico said something about making a mistake parsing the tracker packet. I hope that's at least a small lead.

by **Drakona** » Wed Oct 30, 2013 10:15 pm

Oooh . . . very helpful.

I think that's almost enough to let me figure it out, but I'm too tired to think straight. Tomorrow.