On Fri 01 May 2009 at 07:28AM, Danek Duvall wrote:
> On Fri, May 01, 2009 at 03:18:16PM +0100, Darren J Moffat wrote:
>
> > Given you already depend on OpenSSL anyway why is it preferable to use the
> > (very likely slower) Python hashlib version than a highly optimised (using
> > appropriate cpu asm code) OpenSSL version ?
>
> hashlib has a back-end that links against libcrypto, and uses that if it's
> been built, but has its own fallbacks otherwise. So by using hashlib, we
> don't have a direct dependency on openssl, but will use it if we can. This
> makes the pkg(5) code more portable, but still get the performance we want
> on our core platforms.
It doesn't seem to me to be too unreasonable to think that we could
have our own hashing veneer layer which (for now) uses OpenSSL (or
if not present perhaps supplies its own naive C version), and
later uses hashlib when we get to 2.6.
That seems like something Darren could work on as a unit-testable
piece.
Throwing Python 2.6 in the way of other goals seems like something
to be avoided if possible.
-dp
--
Daniel Price, Solaris Kernel Engineering http://blogs.sun.com/dp
_______________________________________________
pkg-discuss mailing list
[email protected]
http://mail.opensolaris.org/mailman/listinfo/pkg-discuss
