Re: [lwip-devel] [lwip-commits] [SCM] UNNAMED PROJECT branch, ppp-new, updated. c268c5e07c046eb2cb8e5798a1f3eba7e0ad13c1

[Sylvain Rochet]

Hi Simon,


On Mon, May 21, 2012 at 09:23:44AM +0200, Simon Goldschmidt wrote:
> 
> I just now found the time to have a look at the new "ppp-new" branch 
> and found you have deleted all encription code. I'd be in favour to 
> leave it in to make future development easier.

Yup, as said in the other thread, will do, with macro switchs between 
bloated-included-DES-only-usable-for-tests and lovely-PolarSSL-but-GPL :-)

> (We just would need to replace the files des.h and des.c, wouldn't 
> we?)

That would be lovely, but it's not that simple...

We have a 3x56 bits input from the MD4 output. (Actually 56 + 56 + 16
+ 40 padding with 0).

OpenBSD/POSIX setkey()/encrypt() functions uses 64 bytes, yes bytes, 
arrays containing 0 and 1. This is the main reason of its overall 
bloatness.

PolarSSL is clean and uses 64 bits arrays, so we only need to expand the 
56 bits keys to 64 bits ones (56 + 8 parity).

This is why pppcrypt.[ch] expand or collapse the key depending of the 
library used.

Maybe we should extract the DES from OpenSSL, which might be better, and 
seems to fulfill our license needs:

  http://www.openssl.org/source/license.html
  "Actually both licenses are BSD-style Open Source licenses."

But, well, erm, to be honest, I wanted to avoid putting my hands into
OpenSSL :-)


> Other than that, thanks for the great work!

Yeah, this is kind of an amazing stuff to do :-)

I wanted to avoid modifying the pppd code too much but I "failed". With 
more experience and data I will explain why this is just an impossible 
target to achieve.


Sylvain

signature.asc
Description: Digital signature