Re: Proposal: Addition of a random number generator to GLib

[Jeff Garzik <jgarzik pobox com>]

On Sun, 21 Mar 1999, Josh MacDonald wrote:
> Quoting Jeff Garzik (jgarzik@pobox.com):
> > On Sat, 20 Mar 1999, Josh MacDonald wrote:
> > > Quoting Jeff Garzik (jgarzik@pobox.com):
> > > > Otherwise looks good.  This will nicely complement the checksum module I
> > > > would like to add to Glib, for security work and authentication
> > > > especially.
> > 
> > > If people are interested I can add SHA-1 and MD5 into glib right away.
> > 
> > I am interested in developing a checksum API, much like the random
> > number API being proposed.  The user can pick SHA, MD5, polynomial
> > hash, or a user-provided routine.  And there are functions that require
> > a GChecksum structure to maintain state, and other functions that
> > directly create/update the checksum w/o requiring state variables.
> > Attached is the MD5 impl I was looking into, which benchmarks faster
> > than the reference code.
> 
> Second reply.  I'm concerned that glib is creeping out of focus.  I am
> currently using openssl for all my random number generation, checksum,
> and encryption needs.  I also use checksums in a lot of places, and 
> I don't want everything I use to depend on openssl.  Adding a few checksums
> into glib would allow me to remove the code from Xdelta and PRCS2, but
> your proposal duplicates a lot of effort that people have put into these
> interfaces in other packages.  What does everyone else think?

Haven't seen any replies...  What exactly is Glib's focus?

For the checksum API, I figured it would be ok since we have the primes
module already, which appears to be used in one only project in GNOME
CVS, your xdelta :)

FWIW, I use checksumming in one form or another very frequently, mostly
in applications totally unrelated to encryption and encoding.  I think
having MD5 and CRC (at least) in Glib would be very useful.

	Jeff