On Fri Jun 08 2007 at 11:00:35 +0100, bifferos wrote:
> 
> --- David Young <dyoung@pobox.com> wrote:
> 
> > On Wed, Jun 06, 2007 at 06:37:19AM +0300, Antti Kantee wrote:
> > > On Tue Jun 05 2007 at 10:25:41 -0400, Allen Briggs wrote:
> > > > > * tiny & full tcp/ip stack (like QNX for example)
> > > > 
> > > > If we're defining a "tiny & full" tcp/ip stack, it would be useful
> > > > to identify what "tiny" and "full" mean.  Can you elaborate?  I
> > > > have some idea what I would mean by that, but would like to hear
> > > > what you would be willing to give up.
> > > 
> > > One option for a tinier & "full" tcp/ip stack is to rip out the routing
> > > code.  But it's not exactly easy make it an optional module.
> > > 
> > > see:
> > > http://www.cs.hut.fi/~pooka/pubs/EuroBSDCon2006/bsd_lwrouting.pdf
> > 
> > I believe this talk about the size of the IP stack is a distraction,
> > and we should not squander our volunteer developers' valuable time
> > discussing it any further, when we know that people eschew NetBSD for
> > embedded use more often because NetBSD lacks board/device support and
> > essential features such as a NAND flash filesystem.

You can't force volunteers to either discuss or not discuss it ;)

But on a more serious note, the point is more in making the networking
stack more maintainable than in making it small.  Optionally tiny is
ideally just a side effect of "maintainable".

The paper does describe a very real case where the size of the networking
stack was the showstopper ...

> I agree.  And with few 'micro' alternatives to the standard network 
> utilities (dhclient etc..) that is probably where the bloat comes in.

... however, NetBSD has lots and lots of easier fruit to pick if starting
from a stock distribution.

I will now stop squandering our/your valuable time.

-- 
Antti Kantee <pooka@iki.fi>                     Of course he runs NetBSD
http://www.iki.fi/pooka/                          http://www.NetBSD.org/
    "la qualité la plus indispensable du cuisinier est l'exactitude"