New Python block cipher API, comments wanted

Paul Rubin phr-n2003b at NOSPAMnightsong.com
Wed Jan 29 02:08:03 EST 2003


Erik Max Francis <max at alcyone.com> writes:
> > So that applications can use it without having to wait for 2.3.1 or
> > whatever.
> 
> Why the rush to even get it into a 2.3 dot release?  Who needs this
> functionality so urgently?

I'd like it.  The author of mxCrypto just posted that he sees a need
for it.  The other folks on the Python crypto list want it.  There is
maybe a larger group (perhaps including you) who don't care about it,
but in that case, why the concern?  The people who care about it seem
satisfied with what's going on.

> If you want it to get in, take the slow route.  Submit (or revise) a
> PEP, and take your time.  You're in a big hurry to get it into the
> version of Python that's already gone alpha, but I don't think anybody
> else sees the point of the urgency.

You seem concerned that the API may not otherwise get enough
discussion before being shipped, but it's gotten plenty on the Python
crypto list, which is where most of the people who care about this
stuff hang out.  If you're interested in it, why not join the list?
Really, when I posted to c.l.py, I thought I was simply inviting some
possible last minute technical comments from people who weren't on the
list.  Instead, I'm getting a bunch of nontechnical metadiscussion
that's just a pain.

As for PEPs, I just looked at the existing ones; very few existing
PEPs are about library modules, and hardly any existing library
modules are described in PEPs.  Almost all the PEPs are about issues
that affect Python internals.  Even the ones that are about libraries
(like 272) seem mostly aimed at module implementers, not users.

Library modules have usually been described for users in ordinary
documentation, not PEPs.  Why do you want to suddenly start using PEPs
in a way they haven't normally been used?

Also, as I mentioned to David, there's already been a PEP for block
ciphers (PEP 272) and it wasn't all that helpful.  I don't think
anyone has ever implemented what it specifies.  Doing another one
doesn't sound terribly worthwhile.  It's just going around in circles.




More information about the Python-list mailing list