Subject: Re: Considering port to new processor
To: Jonathan Stone <jonathan@DSG.Stanford.EDU>
From: Chuck McManis <cmcmanis@mcmanis.com>
List: tech-ports
Date: 03/30/2001 17:12:36
At 04:53 PM 3/30/2001 -0800, Jonathan Stone wrote:
>You *can't* run NetBSD on a PDP-11. Go look at 2.11BSD and why Steve
>Schultz stopped there.  Not all of 4.4BSD would fit into 11 address
>spaces, let alone what NetBSD has grown into since. NFS?  IPv6?
>mmap() and huge shared libraries with only eight 8k "pages"?

I know. And I acknowledge your point. There is an interesting but perhaps 
to this list irrelevant side discussion which starts with "What is 
'NetBSD'" and ends with a list of components that are required to achieve 
'NetBSD'ness. While at Sun I worked for a short while on the SPARC "ABI" 
and we ran into this discussion pretty much head on. For some, IPV6 would 
not be a 'requirement' (or perhaps even an IP stack) for something to be 
'NetBSD', for others, the "snapshot" in all its hugeness defines what "is" 
NetBSD.

The actual value proposition is about leverage. Can I leverage the kernel 
and base libraries in NetBSD enough to produce an operating system that 
runs on a PDP-11 and on which I can compile and run applications that are 
nominally "ported" to NetBSD?

Steve stopped at 2.11 because BSD had moved on to larger architectures (the 
VAX) and they consciously did not look back. Look at uCLinux and you will 
see an operating system that is "enough Linux" that someone using it knows 
its heritage and it leverages important parts of the Linux kernel.

Is possible to go off and create uCBSD and nominally keep the command line 
look and feel and the system libraries the same, but rewrite the guts from 
scratch with its own kernel architecture and its own way of doing things? 
Of course it is, but that has no leverage.

--Chuck