In message <m105xRI-000g6eC@most.weird.com>Greg A. Woods writes
>[ On Thu, January 28, 1999 at 19:24:09 (+1100), Darren Reed wrote: ]
>> Subject: Re: ip_flow.c
>>
>> Or on the other hand, how many people (besides Perry) would actually
>> enable checksumming on hosts doing gatewaying for packets being forwarded
>> given the pros and cons as discussed to date ?
>
>I think I would enable checksums on my host-based routers too.
>
>I think I'd prefer a compile-time switch though.

My impression is that you can buy any number of very cheap routers 
that will vastly out-perform any general purpose system like NetBSD.
I think the ipflow stuff is cool, and I'm happy to see it in the
kernel, but I'd really like to see us concentrate on having a
picture-perfect RFC-compliant forwarding engine, rather than 
try a whole lot of tricks to make it very fast.  We just can't
win that way -- if I want to route OC-3 ATM, chances are, I can afford
a router to do it.  Whereas I'm definitely going to use NetBSD and old 
hardware system as a front end for my 384Kb ADSL line...

I'd love to hear conflicting opinions (and I wasn't yet convinced
that incremental checksum updating violated RFC1812, anyway).

Thanks,

Kevin
kml@nas.nasa.gov