Niklas responds:
> > It has been *claimed* that a CTM distribution of the NetBSD source
> > tree violates some of the license terms of files in that tree.

> I cannot see how CTM can be more of a license-violating method than
> any other packaging.  True, CTM mails diffs, and thus the distribution
> time scope is a bit extended but it still contains all the information,
> starting with a baseline, including all the diffs up to now.

The question might be whether one could use CTM to receive pieces of files
without ever bothering to receive the copyright notice.  CTM also "feels"
more like many sessions sending file pieces than an extremely extended
single transfer (see below).

> That's no more violating that TCP sessions ship individual packets with
> source without the full license text in each IP packet.

The BSD license (which I happened to check first) says that redistributions
must "retain" the copyright notice, not that it must accompany every byte.
Assuming you've read any books lately, you may notice that the copyright
notice is only printed once, not on every page.  (Though ACM and IEEE
magazines print the copyright before each article, since they anticipate
articles being reproduced independantly.)

> packaging at another layer, in my mind.  What about the reget command
> in ftp servers, should that be disabled too?

If people start routinely avoiding the copyright notices by "re"getting
starting at byte 1900 so they can claim they didn't know what the copyright
terms were, maybe it would have to be.

> However I don't think the law mentions anything explicit about output
> from diff(1) used as transfer method of full files.

Copyright law certainly doesn't including the string "diff(1)", but it
certainly says many things, both explicit and vague, about excerpts and
derivative works, and unfortunately copyright law is structured around
creating intellectual property rights for people who wish to prevent
pretty much what CTM does.

If all files in the tree had the same copyright notice, it would probably
suffice for a copy of the copyright to be sent in the mail along with the
diff listings (yet another reason to pursue assigning copyrights to TNF,
though you'd still be stuck with the UCB Regents and the Unix Systems
Laboratories copyrights).

> If this is the law, then we might as well stop using software altogether.

Some days I'm not sure this would be a bad thing...