Subject: Re: Am I wasting my time even thinking this way?
To: Luke Diamand <firstname.lastname@example.org>
From: David Brownlee <email@example.com>
Date: 06/21/2000 12:01:50
On Tue, 20 Jun 2000, Luke Diamand wrote:
> I wonder if I could ask people's advice about the NetBSD IP stack,
> The company I work for (Virata) develop a range of products (mainly DSL
> modems) that need an IP stack. So many years ago someone wrote one.
> But now we find that our aged and trusty IP stack is no longer up to the
> job. We would dearly like a number of features we really do not wish to
> implement ourselves.
> Obviously the easy solution is to go to someone like RouterWare and buy
> their (no doubt excellent) IP stack. This is very tempting.
> But an alternative would be to try to port the NetBSD stack to our
> proprietary operating system (no - before you ask, switching to NetBSD
> in toto is simply too daunting to consider).
> * Is this a totally evil thing to do in the first place?
> * Will we struggle on an ARM7? If we don't have an MMU?
> * Will we get good performance (who wants a slow router?)
> * Will we struggle to disentangle the IP stack from the OS?
Someone ported NetBSD's TCP/IP stack to VxWorks a while back
so you may be able to leverage off their work, or maybe even
hire them for the job :) That was pre IPv6.
Another option might be to try interesting one of the consultants
listed on the NetBSD website in the job.
You don't state what your clock speed on the ARM is - the NetBSD
TCP/IP stack runs fine on machines substantially lower than the
If you had an MMU I would have strongly suggested evaluating
NetBSD on the box, if only to give a good reference platform
which could be easily kept up to date with further networking
improvements (not to mention a system that can be run from
flash with no nasty licenencing conditions or fees, and a good
availability of expertise :)
-- www.netbsd.org: No hype required --