Subject: Re: Questions about features of NetBSD
To: J.T. Conklin <jconklin@netcom.com>
From: Greg A. Woods <woods@kuma.web.net>
List: current-users
Date: 04/11/1995 09:11:46
[ On Mon, April 10, 1995 at 17:03:02 (-0700), J. T. Conklin wrote: ]
> Subject: Re: Questions about features of NetBSD
>
> I feel that the fat binary concept doesn't scale well beyond two
> architectures, and NetBSD currently has 7 (i386, m68k4k, m68k8k,
> sparc, mips, ns32k, alpha) with more on the way.  And to support this
> well, you would need cross toolchains for each target (not a bad idea
> in general, but I wouldn't want to ship them by default).

Might be worth taking a closer look at how Bell Lab's Plan 9 does this.

FYI, in my no doubt over-simplified way of giving a summary, they use
something akin to union mounts to lay the architecture dependent /bin
over the generic /bin.  The Plan 9 compilation tools also try their best
to build binaries for *all* supported platforms in one go (i.e. all the
tools are set up to do this for you).  This latter bit is what I
consider *most* important!  [Note that it's not so bad as you might
think since they're obviously not using anything with anywhere near the
overhead of GCC and friends.  They can recompile the whole O/S and
tool-set on a decent machine in a few hours (if memory serves me
right).]

-- 
							Greg A. Woods

+1 416 443-1734			VE3TCP			robohack!woods
Planix, Inc. <woods@planix.com>; Secrets Of The Weird <woods@weird.com>