How would you deal with the performance vs. IP tradeoff?

I'm surprised there hasn't actually been a post here about this issue, which seems to be mentioned on some blogs and on the SL Dev mailing list.

The essential problem comes from these facts:
* Running program code that protects IP will take processor time and memory. This will reduce SL's performance.
* Removing even the current, very limited, methods that SL uses to protect IP would increase its performance.
* No security in SL can be secure forever because the SL viewer is open source, and SL lacks the political/financial influence to demand end-to-end hardware security in the same way movie studios do.
* However, most users aren't very technical and the security being breakable with reference to the source code would only make it easier for a relatively small number of people.
* Increasing performance could attract more customers to stay in SL who would buy things, thus increasing sales. However, decreasing security would make content theft easier and the extra customers would join the market for the thieves, too.
* Increasing security could prevent content theft and thus increase sales and security for creators. However, the thieves might just go without content they can't steal rather than buying it, so sales might not go up. Moreover, it would decrease performance and potentially put off people who might otherwise have become paying customers.

So
a) what is your own opinion on the tradeoffs involved in terms of how SL should be implemented?
b) can you think of any way this can be dealt with without somebody having to impose their decision on somebody else who might not agree?

there is no significant performance hit for the type of security being asked for. A few bits need to be checked when an item is copied. It's not something being tested on a per-frame basis, such as physics. "performance" is controlled by factors far more significant than copy protection.

You didn't mention which blogs, but the discussion on SLdev is about the performance of the cache specifically, not SL in general.

The basic question is: how much do you want to obfuscate the cached textures given that they're trivial to retrieve as it is.

Anything that doesn't store them as an as-is usuable format (ie TGA) is probably enough of a deterrent already. Anything more wouldn't make any difference.