Netcode and other questions

7 replies [Last post]
fourier
Offline
Joined: Jan 2011
Posts:

Hello,
I have a question regarding the unlagged implementation. Looking at the credits, I assume you're using haste's unlagged code as the base for your alterations (adding unlagged projectiles and so on), but I haven't seen anything about which version. Unlagged 2.01 was server-client delagging (client side prediction optimizations). I know consistency problems have always existed with the older server-side only delagging. Obviously if the client doesn't have the mod, you're either not delagging the client or doing server-side only delagging; so If the client does have the mod, do you do the full server-client delagging or just server-side?

How much did you alter haste's hitscan delagging code? He did an excellent, simple, and consistent implementation of delagging hitscan weapons, so I just want to find out if there were alterations which may potentially end up being detrimental to consistency (in my opinion only).

I've just discovered this mod, and it looks like it's got a lot of features, so kudos there. One thing I'd like to see in here would be the pmove_accurate code (similar to pmove_fixed, but clean, smooth, and not buggy), which I could provide if interested.

D e c e m b e r
Developer
brb24hours's picture
Offline
Joined: Aug 2008
Posts:
Re: Netcode and other questions

Please wait while I put myself deep into Neil “haste” treatise and then I could answer (since I am not the one who did implement the unlagged)

.DoK. Crazy
Forum moderator
lilcrazy.style's picture
Offline
Joined: Apr 2009
Posts:
Re: Netcode and other questions

If this stuff would means that when i move my model right, the hitbox will move together the body(and not where the body isn't anymore), in a more accurate way, than Nice! and wish you GL in finding a solution asap Winking !

And mb the fake hit-thru could stop to annoy too Straight face . I hope so Happy .

My ideas and content for E+:

.DoK. Clan® founder/leader since October 2008 ©.

.DoK. Clan Trailer <-- Thanks to DoK SoftAir.
Throphy: 3rd place AFC 2010.
]DOK[ DanGeR FroSteD! 95.130.10.38:27960

Caipiroska-fragola , Sex on the beach, Pink floid Big grin
Soft prendiamo i soliti? ok, 1 canarino e 3 ACE con ghiaccio Big grin

rage+
Forum moderator
Zack's picture
Offline
Joined: Sep 2003
Posts:
Re: Netcode and other questions
.DoK. Crazy wrote:

If this stuff would means that when i move my model right, the hitbox will move together the body(and not where the body isn't anymore), in a more accurate way, than Nice! and wish you GL in finding a solution asap Winking !

And mb the fake hit-thru could stop to annoy too Straight face . I hope so Happy .

isn't this to do with timenudge?

fourier
Offline
Joined: Jan 2011
Posts:
Re: Netcode and other questions

I think a negative timenudge would create the affect he seems to be talking about -- with unlagged code, I don't care if people like to negative timenudge so long as they don't complain about the visual inconsistencies. In baseq3 netcode mods, I think it should be banned (and was at least enforced back in the osp days). It's a disgusting hack that creates inconsistencies on both ends; inconsistencies on the client's end: acceptable; inconsistencies on the other clients' views of the game: unacceptable.

Anyway, Crazy, if your comment was in reference to the last thing I said about pmove_accurate, then the answer is that it doesn't have anything to do with that. Pmove accurate mimics pmove_fixed which was intended to produce consistent movement (i.e. not frame-dependent). pmove_fixed has all sorts of bugs associated with it, chiefly the hit detection get screwed up at times. That's why it should never be enabled on anything but a trick server. pmove accurate gives you the smoothed "125fps" pmove_fixed speed (125 and 125 pmove_fixed are different -- the latter accrues slightly higher errors on all axes) without the hit detection bugs and weird other bugs (the overbounce stuff for instance).

It's a client cvar that must be sent to the server as well so it can run the same prediction code. It won't be any different in hit detection -- it may appear smoother but your hitbox will still be in relatively the same location to wherever your model is being drawn.

Am I right in saying there is no command to disable fallkick and damagekick in excessiveplus?

.DoK. Crazy
Forum moderator
lilcrazy.style's picture
Offline
Joined: Apr 2009
Posts:
Re: Netcode and other questions

Well, yes I am 50% in reference with that, now that i understand a bit better.
I always used cl_timenudge 0, negative isn't fair.
OT: :\ here my unlucky list if u're curious, or, just for fun
unlucky1, unlucky2, unlucky3 .

Anyway, i don't wanna add other here, i'm not made for dev things Big grin
regards

My ideas and content for E+:

.DoK. Clan® founder/leader since October 2008 ©.

.DoK. Clan Trailer <-- Thanks to DoK SoftAir.
Throphy: 3rd place AFC 2010.
]DOK[ DanGeR FroSteD! 95.130.10.38:27960

Caipiroska-fragola , Sex on the beach, Pink floid Big grin
Soft prendiamo i soliti? ok, 1 canarino e 3 ACE con ghiaccio Big grin

[MR.]MIRCWAR
tyfon's picture
Offline
Joined: Mar 2005
Posts:
SE Sweden
Re: Netcode and other questions

Main devs are pretty much inactive, have a look through the feedback forum and it's archive part for some clues Happy (and some bug whine....)

GrosBedo
grosbedo's picture
Offline
Joined: May 2010
Posts:
Re: Netcode and other questions

I don't know if pmove_accurate was implemented following this discussion, but while it works well, it's not simulating exactly 125 Hz physics since there are some tricks that are impossible when this cvar is activated.
See my post here:

http://www.excessiveplus.net/release/excessive-plus-v2-2b?page=22#comment-376524