[sldev] Cache politics: performance vs obfuscation
Dale Mahalko
dmahalko at gmail.com
Thu Jun 12 01:57:53 PDT 2008
On Wed, Jun 11, 2008 at 4:09 PM, Thordain Curtis <thordain at thordain.com> wrote:
> As far as the proposal for implementing asymmetric encryption for assets...
> It's a moot point until a major graphics chipset vendors starts working with
> it. Even if Linden Labs were to convince NVidia to create such
> functionality, it would likely be years before it appeared on the market,
> and nearly a decade before LL could say "OK....we're turning off support for
> non-secure cards".
This market segment is in its infancy and there is still time for
technologies to develop. At one time people wondered if specialized 3D
cards would have any market at all, and now they are practically
essential. In terms of deployment time, the big market players are
able to bide their time, for years if necessary, for their DRM to get
out into the marketplace.
Did you know those SD memory cards contain a DRM content management
system? SD doesn't mean "SanDisk".. it means "Secure Digital", but
their DRM is not being used for anything yet, and probably will stay
dormant until SD becomes the de-facto memory card standard. We seem to
be moving in that direction, with SD winning out over most other
formats.
> Not to mention the asset server would have to custom
> encrypt every single texture request with the client's graphics card's
> public key. This would increase the asset server's work load by an order of
> magnitude, and last time I checked ..... we don't want that.
Heh that is not how most DRM works. Distributed media using
mass-market DRM has all the same encoding, so the devices share the
same private keys and all use the same public keys. So there's no need
to encode data to match just one 3D card. The data would be encoded
once and work across all 3D cards using the same public keys.
Having a different public/private key for each DRM device? That is
possible but would require the stream be specifically encoded to match
that device, and as you say it would be several orders of magnitude
more work for the asset servers. I am only talking about the first
type, where the data would be encrypted once and sent to all users in
the same pre-encrypted state.
- Scalar Tardis / Dale Mahalko
More information about the SLDev
mailing list