Skip site navigation (1)Skip section navigation (2)
Date:      Wed, 5 Apr 1995 06:14:11 -0400 (EDT)
From:      Peter Dufault <dufault@hda.com>
To:        terry@cs.weber.edu (Terry Lambert)
Cc:        nate@trout.sri.MT.net, rkw@dataplex.net, current@FreeBSD.org
Subject:   Re: cvs commit: src/sys/i386/conf Makefile.i386
Message-ID:  <199504051014.GAA03887@hda.com>
In-Reply-To: <9504042045.AA18525@cs.weber.edu> from "Terry Lambert" at Apr 4, 95 02:45:55 pm

next in thread | previous in thread | raw e-mail | index | archive | help
Terry Lambert writes:
> 
> > *BLECH*  BSD systems have *never* (and should *never*) require that you
> > have the complete source tree installed just to build a kernel.  Making
> > them go through alot of trouble to build a kernel is a waste of their
> > time.  Many, many more people build kernels w/out src trees than people
> > who build kernels w/src trees.  We are trying to make the system
> > *easier* to use for the avg. user w/out penalizing the developer.
> 
> I'm glad you said this.
> 
> When will the binary link kit be available?
> 
> I've often thought that it should be possible to configure devices
> in and out of the kernel without regenerating more than a single
> object file containing the device switches.

Wait a minute, who was that Terry Lambert giving me such grief about
this last week?   I thought we had to redesign the configuration, power
management, and make the kernel multithreaded (no, that last
one was a different issuse, never mind) at the same time.

At the moment you can almost do this for isa drivers except for
the assignment of interrupts in isa/isa.c.  I just steal
unused interrupts to get around the problem.

> 
> The really nice thing about this, of course, is that multiport
> board drivers and Adaptec drivers using Adaptec code, and (potentially)
> the Intel supplied-for-MACH binary math coprocesser emulator could
> all be supplied without sources in full conformance to the non
> disclosure agreements required to obtain them.

I agree with this completely.  Fundamental changes to company policy on
intellectual property are best left to the Linux camp.

> 
> To make the nay-sayers happy, putting the source in escrow with the
> condition that a non-disclosure be signed to obtain it should be
> sufficient to not vest too much in a driver writer as a single point
> of failure.

I don't know about this.  You'll have to be very careful about who these
sources go to.  One slip and FreeBSD has a black eye and a lawsuit.

> That's just a legal niceity, however, and the main point, that it be
> *possible*, is the important thing.
> 
> Actually, sufficient reliance on a devfs, loadable modules, and
> soft autoconfiguration with a "save" for "/kernel -c" boot should
> make it totally unnecessary to have a link kit to get a new kernel
> configuration anyway.

Yes, this is a good design rule and with a little attention to
detail it won't even be much work.

-- 
Peter Dufault               Real Time Machine Control and Simulation
HD Associates, Inc.         Voice: 508 433 6936
dufault@hda.com             Fax:   508 433 5267



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199504051014.GAA03887>