[haiku-development] Re: Design for signed packages
- From: Jonathan Schleifer <js-haiku-development@xxxxxxxxxxx>
- To: haiku-development@xxxxxxxxxxxxx
- Date: Sun, 23 Mar 2014 03:11:13 +0100
Am 22.03.2014 um 22:55 schrieb Jonathan Schleifer
<js-haiku-development@xxxxxxxxxxx>:
> Looking some more at it: It copies the whole message, so either we need to
> change the reference implementation to not do that, or use my original
> approach and hash first. The message it signs is of variable length and I
> could not find a maximum length, so in theory, it should work if we patch it
> to not copy the message. I'll try to contact djb and ask him what he thinks
> is the right approach.
Of course, hashing the data before passing it to Ed25519 turned out to be a
brain fart. I had a deeper look into the implementation now, and my impression
that hashing beforehand is unnecessary was right: get_hram() basically does
SHA512(r, pk, m) and in the place where it matters it was SHA512(extsk[32],
..., extsk[63], m). So, that copying-behaviour was just bad API design in the
reference implementation.
I decided to change that in now to not copy the message, so please, everybody
review my patchset:
https://github.com/Midar/haiku/commit/a9d39808cca3c28e14d99eabeb624a3207331d87
It also becomes obvious in my change that hashing beforehand is absolutely
unnecessary ;).
--
Jonathan
Other related posts:
