[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Public WebGL] Microsoft Weighs in on WebGL Security Issue

I agree with Chris.


On Thu, Jun 23, 2011 at 7:17 AM, Chris Marrin <cmarrin@apple.com> wrote:

On Jun 21, 2011, at 2:33 PM, Patrick Baggett wrote:

> They claim to have fixed it in Silverlight 5 RTM. So retest when it comes out and if it isn't fixed (which it probably won't be) then reopen/refile. Hold them accountable for what they claim to have done, which is prevent DoS attacks using the GPU despite GPU/drivers not having explicit support for this. If they have, then they are a step ahead of the competing technologies and should be praised. If they haven't, then they are quite clearly lying (considering they have a test case that reproduces the problem 100% of the time).

It's very possible to avoid DoS attacks with proper handling of GPU reset. This requires support in the graphics driver and I'm sure some do a better job than others. So it's important to test on a variety of drivers. But the bigger issue for Apple is the handling of "innocent victims". What happens to other apps using the driver when the guilty app causes a DoS? DirectX supposedly has a way to inform the innocent apps so they can clean themselves up properly.

So the more interesting test, IMHO, is to have a windowed Direct3D based game or two running in another window. Get them to an interesting state (running in the middle of a level) and then generate the DoS in Silverlight running in a browser. What happens to the games? Are they unaffected? Do they reset to the start of the level? Do they crash? Any of these would be an indication to me that the bug is surely not fixed. Furthermore, it's a bug that can't be fixed in Silverlight. It has to be fixed in the game or the driver. If one game is found with a problem then you could say this is simply a bug in the game. But if it happens in many, then it's really a system problem.

This is why Apple needs a driver solution that avoids innocent victims. Not only are OpenGL apps written without the concept of having to handle the innocent victim case. But even if they were, it would often be difficult or impossible to deal with a GPU reset in a clean and consistent way. It really needs to be the driver that prevents the case from ever coming up.


You are currently subscribed to public_webgl@khronos.org.
To unsubscribe, send an email to majordomo@khronos.org with
the following command in the body of your email:
unsubscribe public_webgl

Tony Parisi                             tparisi@gmail.com
CTO at Large                         415.902.8002
Skype                                     auradeluxe
Follow me on Twitter!             http://twitter.com/auradeluxe
Read my blog at                     http://www.tonyparisi.com/