Code review comment for lp:~vanvugt/mir/nonprime-server-support

It's not a security question. It's a question of having no choice other than hacking the kernel driver (and then waiting for all existing users to then get a new kernel before we can support them). But given the more secure feature support, that is still used as first preference here.

I already pondered the idea of having a different implementation of Buffer. Decided that would be inappropriate because even the flink code path uses GBM. So both code paths are GBM code paths. If you were to separate them then even the existing one would need renaming from GBMBuffer to PrimeGBMBuffer. But I don't like the idea of having duplicated code. I guess it would have to be prototyped to see if it's just an acceptable level of duplication.

Personally I would still like to keep a single class "GBMBuffer". That's short and sweet and the most accurate representation of what it is.