ken Powell - Sun Microsystem wrote:
> James Carlson wrote:
>> The latter would make it possible for you to set PARAM_IGNORE_LIST so
>> that the DHCP client would not use that option.  (It's not a fix, but
>> it'd be a useful work-around, just in case there's someone out there
>> with a DHCP server that gives out bogus netmasks that aren't "obvious.")
>>
> 
> Just ignoring the netmask would still fail at the hotel I was at.
> The default netmask for address 192.168.7.216 is 255.255.255.0.
> I needed a netmask of 255.255.252.0 to reach the router.

It'd be better than what we have now.  I believe what happens now is
that we set the netmask and then either fail (because the ioctl fails)
or explode later when you try to change it to be right.  At least with
an ignore option, you could potentially move this one thing to the
user's hands.  (I guess that adds a third thing to my list: if netmask
is ignored, then the abandon logic needs to ignore it.)

I wonder how Windows manages to work on that network.  If they're given
a next hop address for a default router, and that address is off of the
local network, do they just ARP for it on spec?  If so -- and it's the
only way I can see this working -- that's bizarre, and would likely be
hard to emulate.

Maybe the best we could do would be to detect that the default router
address we've been given is off of the local network specified, and --
within some constraints -- just widen out the configured netmask so that
the router becomes reachable.  It's certainly a "wrong" thing to do, but
far less wrong than teaching ARP to ignore subnets.

I've often wanted a "you and the horse you rode in on" option for a
return message to peers that violate the protocols this badly.  This
situation would be a good application.

-- 
James Carlson         42.703N 71.076W         <[email protected]>
_______________________________________________
networking-discuss mailing list
[email protected]

Reply via email to