MD4 bug-fix for protocol version 27

[Donovan Baarda]

On Thu, 2003-05-08 at 08:56, jw schultz wrote:
> On Wed, May 07, 2003 at 10:26:57AM -0700, Wayne Davison wrote:
> > On Thu, May 08, 2003 at 02:39:57AM +1000, Donovan Baarda wrote:
[...]
> OK.  Two issues here.
> 
> 1. Is this fix correct to give us full md4 compliance?
>    Craig?

[phew, not me]

I haven't looked at the rsync md4 code for a while (hence not realising
the fixes were in CVS). I'm burned out on optimising md4 for libmd and
librsync... I'm not really interested in looking at yet another
implementation. I would really encourage you to adopt the libmd API so
that you can offload the md4 code, but I'm too slack to figure out the
librsync code and submit patches, so take my advice with a grain of salt
:-)

> 2. If so, do we bump to protocol 28?  It has been a month
>    since protocol 27.  This would break things for those
>    with cvs versions since then.  I was going to describe it
>    as a rare corner case (length % 64 == 4) but realized
>    that amounts to more than 1%.
>    
>    The issue is whether to break things for those few who
>    are running cvs that would mix cvs versions or to hasten
>    the day when 2.5.6 becomes incompatable.
> 
> Lets hear it.

My attitude would be; cvs is not released, use at own risk. If you try
to maintain backwards compatibility between each minor cvs checkin, you
will tangle yourself into knots. Backwards compatibility for releases is
a big enough burden to bare.

-- 
----------------------------------------------------------------
Donovan Baarda                http://minkirri.apana.org.au/~abo/
----------------------------------------------------------------