Coop multiplayer support?

[poVoq]

In your FAQ it sounds a bit like multiplayer isn't supported due to performance problems with the QC entity system.

But maybe coop multiplayer would still be feasible? A campaign with 2-4 players would probably be really fun.

[Subject9x]

whew boy that's it's own odyssey. I can kind of point to where things need to be fixed, but I ran out of energy to focus on them.

1. Client-side prediction doesn't exactly work right, at best I was using the Darkplaces flag sv_clmovement 1.

2. battleMETAL's QC uses an entity for player camera, an entity that has to follow the player's "true" entity.

3. I never got FTEQW-style entity updating to work right either.
   https://github.com/Subject9x/battleMETAL/blob/master/client/client_main.qc#L597

To really get this codebase back to network play; we'd have to get clientside prediction to work for player cameras, and be able to sync CSQC entity copies to their SSQC counterparts. For me, this was a lot of work, but it's not "impossibly hard".

[Subject9x]

Another random note: Co-Op / PvP was originally a design goal, and there's still some code that functions for this:

• you can change your faction id via console: cmd mech setfaction <1, 2, 3>
• this then allows you to set Deploy and Nav points with different factions.
• You can set player starting equipment on Co-Op maps by making a .sav file in the /data/map_scripts/ folder for that map.
• you can check out local co-op yourself with this fun method:
  • start up 2 instances of darkplaces.
  • instance 1, go to any map, and enter the following console commands:
    • coop 1
    • maxplayers 2
    • restart
  • then on instance 2, type:
    • connect
  • you should be able to connect to the map in instance 1.

Another fatal showstopper bug though is that AI movement is not predicted clientside and apparently has a tendency to chug the server. If network play were to be implemented, the AI would need a semi-complete rewrite of its movement code to ignore engine builtins of walkmove() and movetogoal().

[poVoq]

Ok, thanks for the details. I guess we can keep this open to track possible progress towards this "not impossible" goal in the future :)

[Subject9x]

Agreed, maybe if we can get a build of the game ported over to FTEQW - I've heard that networking in that world is more robust and easier to wrangle. 👍

[Subject9x]

no joke, an idle conversation and some testing has lead to an idea to try for networking.
So far the sacrifice seems to be player shadows and maybe some 'sim' feel with looking around your mech.

[Subject9x]

Yup been down this hole the entire time; lag-compensator branch is probably going to be a fairly massive upgrade to the game, just a warning.

[poVoq]

Keep us posted, sounds like it will improve the game a lot :)

[Subject9x]

Still on-track to finishing up the network branch. The last remaining gubbins are:

• building death animations.
• sending player data to other players.
• refactoring mission lifecycle to handle multiple players.

after this there's still the UIX work to be completed but I can merge the network changes to the main branch and work from there.

[poVoq]

Sounds great. Enemy AI is working so that coop missions would be feasible?

[Subject9x]

From my 2-player tests with a friend, the AI functions more-or-less the same, but they can be a tad network heavy.

[Subject9x]

wrapping up internal tests this week. The lag-compensation branch seems to mostly work but I'd still say keep the player count under 8. I've been testing with 4 players so far and some latency does seem to build up over time but nothing that stops the show.