Ok, Dominic. Thanks for the insight.

Yea, I know fixing forever the cheat problem is a utopia. Or at least
looks like is something hard to achieve. I guess future technology (?)
will able smaller PVS so wallhacks will render much less usefull. Also
cheat players play diferently than normal players, other approach,
analysys statistic is posible (??)

;)


Dominic wrote:

As far as opengl hacks are concerned, the hacked game is executing exactly
the "correct" calls, plus some the game itself doesn't know about.
So even with the scenario from your "perfect idea", the game would still
execute the correct calls, and the hack would add some of it's own, like
disabling depth-testing, coloring models, drawing esp and so on.

What you would have to check is the exact call-path of every opengl
function, and compare that to what a clean machine calls. This is
unfeasible for various reasons:

- It would impose heavy overload on the client - models for example
consist of a lot of vertices, each results in a opengl call

- A hack doesn't necessarily call the opengl functions exported by
opengl32.dll, it may as well call the driver directly, so you'd have to
add trace support for every driver

- Almost all drivers use different implementations of the function calls
for different situations, like first vertex, 2nd and following vertices,
current shade model, and so on, so even if you traced every possible
driver (nvoglnt.dll, ati equivalent, wickedgl, ...), you'd also have to
consider every possible optimization. Nvidia's nvoglnt.dll calls gl
fuctions itself, for example if you issue a glColor after an glBegin (or
some other basic functions, I don't remeber the two I stumbled across some
weeks ago), probably to adjust to the new state of the opengl machine

What I want to point out: it may be (theoretically) a good idea, but it's
never going to work.

Kind regards,

Dominic


On Mon, 24 May 2004, tei wrote:


Here is pic
http://telejano.berlios.de/option/anticheat1.png

the perfect idea is to generate all opengl calls, but store and not draw.
 then send the whole data has a array to the client. the dumby client
will render the whole stuff.

the perfect idea is imposible, has you need infinite bandwidth and zero ping

the un-perfect idea is to generate the data ALSO at clientside, and send
a hash of the data from client to server, then the server will send a
auth code if the hash of the render data on server side mach

the un-perfect idea is ugly because client is fat, server is fat. the
server will need much more horsepower to GENERATE the opengl calls for
all players, but not so fat... will not need a video card, has not need
to draw anithing, only to pretend so.

its also ugly because to hash mach, is need to use the same exact
eyecandy setup on server and client. If the client active "rain", the
server will need to active rain, ..so server is need to be aware of
eyecandy options on client.

What people think about this idea?

--Tei

_______________________________________________ To unsubscribe, edit your list preferences, or view the list archives, please visit: http://list.valvesoftware.com/mailman/listinfo/hlcoders



Reply via email to