Re: [PATCH] cryptoloop

[Christoph Hellwig <hch@infradead.org>]

On Thu, Jul 03, 2003 at 06:25:59PM +0200, Andries.Brouwer@cwi.nl wrote:
> They are used much more frequently than cryptoapi is.
> People tell me jari-loop is much faster at present.
> If this is true, your move would not be very popular.

Personally I care far less about one cryptoloop implementation
beeing faster than another one, but about proper design of whatever
gets into mainline. Think about this like the freeswan vs kernel
ipsec thing - of course klips had more feature and more mature code
initially (and maybe still has in some areas) but the kernel ipsec
is the much better design.

I'd be very interested in seeing some backing and explanation of his
claims so we can incorporate it into the generic cryptoapi / loop code.

The only thing I could imagine is that he has assembly implementations
of many ciphers that are faster than the C ones in cryptoapi - something
that's on the cryptoapi todo list anyway.

> Anyway, I am not doing a redesign. Just a cleanup.

Umm, no.  You add a feature.  That's something very different from
a cleanup.