No feigning :(

[o_lucid]

Yes, it is slightly disappointing that there will no feighing as of right now. Although this probably will change. From what I have been reading about what other people have been describing as engine limitations; couldn't you create a feign by changing the models animation and disabling collision detection. Then alternatively rather than flying like a ragdoll as soon as they were damaged feigning would be disabled.

[o_frenchtoast]

Originally Posted by lucid

Although this probably will change.

Actually, it probably won't. The devs have been pretty straight forward on the issue the whole time.

And as for your idea, it wouldn't work. If you give the feign a specific animation, then it'll be obvious when someone feigns. The point of feigning is that people don't know you're feigning.

[o_ihmhi]

It would be pretty funny if the spy feign were way over the top, like old 1950s Warner Brothers style.

-feign-

"Oh! Ya got me! I can see me dead relatives! I'm comin' Aunt Mable!"

-grabs the collar of a nearby enemy Engi-

"watch ol' Fido for me! I begs ya!"

The the spy falls over and holds a daisy in his hands.

[o_guy incognito]

How did they do the knock down in Oblivion? See... in 3rd person you see yourself fall down in ragdoll then a getting up animation after the body comes to more or less rest.

[o_kam]

that's single player though

[o_dammage]

the whole problem that the ragdoll stuff is local. As soon as someone dies, the server releases it from the multiplayer logic and lets the physics engine work with it.

But if you wish ragdoll feigning, you need to force the server not to release the feigned ragdoll from the multiplayer, but also apply physics engine on it. Not a problem at all, but you want to be able to shoot at feigned spies.

And here comes the problem: it forces the server to compute the physics for the ragdolls. That is a bandwidth and a CPU issue. Fortunately the CPU issue is not an issue at all on all those monster servers.

So, the bandwidth issue comes: you have to transfer the whole trajectory of every single part of the spy to every player. Two options for computing come into mind:
full "uncompressed" computation which is easier to implement and allows 1:1 trajectories on every client. The problem is the information amount produced by such a monster.
The second option could be borrowed from MPEG. Let the server compute the I-Frames (Indexframes) which are full-information snapshots. Those should be computed when a doll hits a surface. And additionally a few P-Frames (Prediction frames) should be also supplied by the server to the clients.
Between all the frames let the clients compute the physics.

That would lower the amount of traffic, but would still not fix the lag problem: the server has to first compute the values and send them all to the clients out-of-band. Else the players would also see a lag between a "local" and a "remote" ragdoll. Out-of-band traffic will have to be hacked into the protocol, too, as far as I know.

http://de.wikipedia.org/wiki/Bild:GOP.gif
Just imagine the B-Frames were away are the time frames where your computer is computing and the I- and P-Frames are the ones given by the server.

The problem is: DOES the SDK allow such hacking?
And the second problem: if the MPEG-approach is used, it has to be so good that even in quiet situations the player cannot notice any artifacts compared to local dolls.

I cannot say what idea the devteam had, maybe something crazy I couldn't think of with my pascalized/ocamlized/c++ized brain

[o_nezumi]

Ok, look, I seem all crazy 'n shit with my size 5 bold text, but not 9 posts later defrag comes in and totally fails to say anything about whether or not feigning will be in the game.

MY theroy is that they're working on some sort of way to get feigning into the game in secret. They don't want to say it's not in the game, 'cause it still may be yet if whatever they're trying pans out (and we all know there are other secret features they're keeping under their hat besides this). At the same time, they don't wan't to say that it is in the game because it'll be a huge let down if it doesn't work.

Regardless, I think there's a cheaper solution dammage. If you could have a ragdoll act like a torso attached to the player when they feign. The torso is physically simulated server side, and stuck to the center (or bottom, whatever) of a box that clips exactly like a ducked player. All the limbs are done client side. The problems with that would be that a limb can't exert forces on the torso, which may look really wierd but maybe passable if the player wasn't moving fast when they feigned, and that the torso can't slide away from wherever the player is sitting while feigned. Players would have to learn to not feign while on steep slopes or near ledges. There's probably some simple reason why this wouldn't work at all that I'm missing, but I'd sure like to know what it is.

[o_dammage]

the problem is that if you do full server-side physics computing while released ragdolls are clientside, there will be a distinguishable lag between the spy ragdoll and the local one - the local one will instantly fly as soon as locally an explosive hits it, the spy doll will first be computed on the server and its trajectory will be then streamed to the clients.

CS:S already has problems with those barrels. It is possible to avoid the remote lag by applying it to the local ragdolls, too - the ping is known, apply techniques similar to quake3's antilag to the local ragdolls, so that the physics is applied to the local one with the same lag as to remote ones.

The second problem are hitboxes. They are dynamic for ragdolls, if you compute the only torso trajectory, the problem arises that if you shoot at locally-computed parts, they could be pretty different on every client's machine as there was no synchronization for them between all clients.

That can be also compensated by applying hitboxes only onto torso of the dolls on all dolls. That would be the easiest approach with the dolls.

The MPEG approach would be the most efficient one. It would require only a quant of the full traffic, but it has issues, too. If the trajectory isn't sent as an out-of-bands stream, but is sent in packet form, any actions which apply to the trajectories during the flight would be flawed. Imagine an explosive throws a doll away and then a second one gives it an another boost.
But that's fixable by using streams, so that the server can just replace old unsent data by new one. If the I/P-Frames are not frequent enough, though, shooting at ragdolls during the B-Frames (local frames) would have no effect, though

I do not know if the sabotage skill reballances the spy. Or better to say: if it removes the need to become cloaked/to feign. If not, I would suggest to implement ETF's cloaking system. It costs cells and takes time to get invisible. Obvious positions to cloak and hitsounds/pain sounds help to reveal a cloaked camper very easily. Soldiers even tend to throw a nail grenade into the flag room just to check the whole perimeter

Nevertheless: I believe the community shouldn't stick to the old and proven too hard. While it lowers the chance to fail, the chance to create anything new and crazy is equal to zero, too. It was just my small evaluation of the possibilities. Beware, it was dangerous half-knowledge

[o_guy incognito]

As far as saying for oblivion, "thats for single player" you know theres a multiplayer mod right.

Its got to be something along the lines of ragdolls on, hell the feign could even be in 3rd person. then the up animation.. I guess I sort of said that before. but arr.

[o_ghost]

The difference is HL2 has been built with multiplayer in mind and Oblivion has not.

I'm not sure you are understanding what's going on when a body ragdolls in a multiplayer game of HL2/CSS. When a player is alive and kicking the server keeps track of that player's position and whatnot. The instance the player dies the server version of the player ceases to exist. A ragdoll version of the player is created on every single client and each client handles the ragdoll on its own. That's why if you ever look at a dead body in CSS from 2 different computers the body may not appear in the same spot on both computers.

In order to make sure the body is in the same spot for all clients the server would then have to handle the ragdoll calculations and pass all that info to every client. This will most likely result in lagging the server when a spy feigns. To see this just pile up a bunch of stuff in HL2M and toss a nade at it. The slowdown is caused by the server having to push all the info about the objects to all the clients.

[o_etzell]

Honestly, in HLDM, there's a slowdown because of all of the objects, I'm thinking that one torso-shaped object isn't a huge deal. Who cares about the limbs, chances are if you're spychecking corpses, you're shooting at the torso ANYWAY...

[o_ghost]

Obviously it must be more complicated than that or else the dev team would be doing it.

[o_zaskar]

Dammage, a difference in look between shooting a dead body and a feigned spy is ok, i mean, if you shot a dead body in the arm or a guy pretending to be dead, the jig is up right there. like in TFC, the spies shot blood when you shot the bodies. Even if the feigned spies diddnt move at all.

Have the spy go invis, and locked to the spot and have the camra moved close to the ground. At that same time spawn a phys prop of whatever he is disguised as and have it fall, but also be pinned to that location. Then just make any shot in a specific radius hurt the spy, you lose headshot abilities on the body but it still is a plan.

[o_ghost]

So instead the player shoots nothing next to a dead body and still produces a hit? Not very consistant at all. Also, you would have no way of knowing what the radius should be and given the circumstances the bodies could be in radically different positions/areas.

[o_dammage]

the problem is not shooting at a body, but hitting it.
Local physics is not synchronous among clients. What your client computes could be very different on an other client's machine.
That can occur very easily. You were not in a magic radius where an explosive came up. Thus the server never sent information about an explosive that came up and therefore your local client never computed physics for ragdolls lying around the explosive.
That is a sacrifice towards less lag and better performance.
Eventually you are shooting (on your screen) at a body and a different player sees you shooting at the bottom a metre away from it.

A more probable situation would be if you join a server "in the middle of the flight". Where should the body land without receiving any data from the server? As far as I remember the HL2 engine just transfers to you positions of static entities.

You cannot just spawn a physic entity on a spy without synchronizing it among the clients and the server. That problem is common to unsynchronized threads seeking the same resource - they all write to it without waiting to each other and you'll be lucky if you didn't develop nuclear plant controllers

Hopefully HL2 physics engine is deterministic. If not, it would be impossible to use it for such purposes (deterministic algorithm = an algorithm returns the same results with the same input data every time. That means that it has no side effects. A side effect could be usage of input/outout or random number generators and other nifties. See it like a mathematical function. E.g. take + - 1+1 always produces 2 and never 3 or 1)

[Circuitous]

Originally Posted by Guy Incognito

As far as saying for oblivion, "thats for single player" you know theres a multiplayer mod right.

Its got to be something along the lines of ragdolls on, hell the feign could even be in 3rd person. then the up animation.. I guess I sort of said that before. but arr.

You know it's a piece of crap and not fully implemented, where the other player is just represented by an NPC in your world, and the synchronization is actually pretty poor?

[o_shoebacca]

Ok, how about spies can do Three Stooges type maneuvers like the One- handed-two-fingered-eye-poke and the Two-handed-eye-poke-defense?

[Hand_O_Death]

I am not thrilled with the picture of the cloaked engineer. It just seems a little too visible. Just a suggestion, but look at the Hidden: Source people. I do not know if that picture is even close to a final cut. I just wanted to throw my opinion out there.

http://mods.moddb.com/5133/hidden-source/

[Circuitous]

He's not that visible when he's actually moving.

On a related note, consider the context. You've hot the Hidden, where invisibility is all this guy has, and you've got the Spy, where invisibility is one of many tricks. They could easily make him entirely invisible, and suddenly playing Spy in a busy server is a fucking cakewalk. If you're not paying attention, he can slip right by - sometimes right in front of you - but if you know what to look for he can be fairly easy to spot.

Invisibility is a handy thing to have but don't expect to be able to rely on it completely and slip by any defense. You won't stand a chance.

[allisvoid]

Originally Posted by Hand_O_Death

I am not thrilled with the picture of the cloaked engineer. It just seems a little too visible. Just a suggestion, but look at the Hidden: Source people. I do not know if that picture is even close to a final cut. I just wanted to throw my opinion out there.

http://mods.moddb.com/5133/hidden-source/

looks fun as hell...!