Host configuration

David W. Hankins David_Hankins at
Tue Feb 19 17:13:56 UTC 2008

On Mon, Feb 18, 2008 at 01:57:47AM +0000, Phil Mayers wrote:
> As far as I can tell, the autoneg == bad meme steps from the bad old days 
> when 100Tx was just getting rolled out and a considerable number of 
> so-called "good" network cards were in fact s**t (e.g. 3c905b). In a 

Well, maybe this ages me.  Before there was autoneg, there was auto-
sense.  Extremely small differences in timing (and curiosities of
implementation) meant two chips of different manufacture could
(in)consistently get the wrong answer.  So the networks in which I am
to blame for their modern 'no auto' policies were formed at a time
when it really was poison.  It doesn't help that before auto-sense we
were going around setting every NIC and switch port manually already,
a thing that often involved entering a BIOS configuration for the
ethernet device (at least on PC hardware).  So the job got done badly
either way, you may as well keep going the old way.  It further didn't
help that although NICs of the era supported full-duplex on the wire,
some still managed only one ring buffer; so they were half-duplex

Some haven't been re-evaluated, some have and are determined to wait,
and others have re-evaluated to reach a new unique position; their
switches and clients are configured to autoneg, but to advertise only
full-duplex varieties.

I realize it's my fault for this OT forray, but I introduced it as
an important allegory; before there was DHCP, there was BOOTP.

Choosing not to select DHCP because of BOOTP-esque qualities I keep
hearing DHCP is supposed to have is just as silly (or not, if you're
truely determined to stick to removing redundancy) as the modern
tendency to deselect autoneg because of auto-sense's legacy.

> different matter. There are in many cases reasons why it doesn't work. The 
> main ones I can think of are servers with >1 IP (e.g. terminating >1 SSL 
> website)

Yeah, that doesn't fit well in DHCPv4.  DHCPv6 _maybe_ but probably
not either.

But if it were me, I'd probably put them on lo0 aliases, then route
the block to the machine.  I'd still use DHCP on the upstream
ethernet interface not just for an address; but for any other config
I can source there.

> and any servers which actually support the DHCP (e.g. DNS).

Again to segregate the idea of ISC DHCP and IETF DHCP; if you use an
appropriate implementation of an RFC2131 client for servers, there is
no bootstrap dependancy, and so no cascade in failure.  Admittedly,
ISC DHCP's dhclient is not one.

Ash bugud-gul durbatuluk agh burzum-ishi krimpatul.
Why settle for the lesser evil?
David W. Hankins	"If you don't do it right the first time,
Software Engineer		     you'll just have to do it again."
Internet Systems Consortium, Inc.		-- Jack T. Hankins

More information about the dhcp-users mailing list