> Excepts from Andrew 
> Excerpts from Excerpts from mail: 12-Jan-95 Re: `use sup' not tar balls.. Ted
> Lemon@vix.com (346)
> 
> > I wonder how useful a daily context diff would be?   It might be small
> > enough that people who care could just have it automagically mailed to
> > them...
> 
> Excerpts from Excerpts from mail: 12-Jan-95 Re: `use sup' not tar balls.. Chris G
> Demetriou@LAGAVU (209)
> 
> > except it's not at all trivial to generate.
> 
> I'll actually agree with cgd on this one :-).
> 
> I think a `pull by demand' approach would be better and safer. 
> Following on from this, the simplest pull approach I'm aware of is
> requesting entire files.  While true, it isn't as efficient as diffs, it
> is much simpler and far more reliable.  Remember the war? (MINIX patches
> :-)

(P.S. Sup works fine for me, but...)

How about tagging the CVS archive when the tarballs are made, and adding a
script to the commit process to take a diff between that tag and the commited
file (could be a background process so it doesn't slow you down too much.)
and dumps it in a directory for retrieval. Anyone wanting to get completely
up to date just gets the current tarballs+diffs, and once you're current
with the diffs you never need tarballs again. Diffs could be ftp'd, or 
possibly sent to a NetBSD-diffs mailing-list.

							David Maxwell
							david@web.net