On Sun, 06 Feb 2000, you wrote:
>
> My thoughts for what they are worth:
> 1) the routing table is set up for multiple subnets and the ports are
> setup for one network.
Do you mean there should be a netmask on host routes? That doesn't make sense.
> 2) I would have expected ax4 to be network 44.131.161.0/255.255.255.0.
But the part of the 44 network that is visible though that port is both
everything (44.0.0.0/255.0.0.0) thru the gateway 44.131.161.230 and other
(smaller) subnets via other gateways
> 3) I would have expected a secondary IP address on ax4 i.e. ax4 to have
> both 44.131.161.242 & 44.131.160.xxx. (Network aliasing) It may be
> that enabling network aliasing is all that is needed, you might not
> need to actually assign a secondary IP address.
Tried aliasing to avoid the outbound route being the same interface/ip address
as the inbound - it doesn't make tuppence worth of difference
> 4) the sorting of the routes looks wrong, would have expected sorting by
> route netmask and then by destination address. Mind you 2.0.36 also
> exhibits the same behaviour, HP-UX & Cisco routers do what I expect,
> can not remember what Solaris and Welfleet routers do (6 years of
> non-use showing).
>
> One of the side effects I would expect to see with the setup you have is
> ICMP redirects. Do you see any?
No redirects at all. Not sure if I have them enabled (defaults?) in the
/proc/sys/net/ipv4/conf/all sysctl stuff
> I would have anticipated IP addresses and netmasks something along these
> lines:
>
> gb7ipd 44.131.161.230/255.255.255.0
> gb7itg 44.131.160.127/255.255.255.0
> gb7va 44.131. y. x/255.255.255.0
>
> the_box ax4:0 44.131.160. x/255.255.255.0
> the_box ax4:1 44.131.161.242/255.255.255.0
> the_box ax4:2 44.131. y. z/255.255.255.0
>
> the_box ax3 44.131.160. x/255.255.255.0
> the_box ax6 44.131. y. z/255.255.255.0
I don't understand this at all - an interface having a netmask? I thought it
was destinations that did. Perhaps you could define the actual route statements
that give this result. Note that I've only got as far as testing the host
routes so I would not expect netmasks to have any effect as I'm only trying to
get to the gateways so far.
>
> However as all my router experience a practice is on cabled networks
> things may have to be done differently on radio. Indeed on cabled systems
> have 3 RX's and 1 TX is something I have never seen, RX's & TX's come in
> pairs.
I would expect you won't see it but I expect that its a common occurance
on RF linked systems - it certainly will be round this way until the quantity of
traffic makes it economic to have a one-to-one RX/TX pairing
> I can not take this any further on a practical note as I am running
> 2.0.36 kernels here and do not have enough gear to put a test rig
> together.
Mind you, 2.0.36 works as I expect - its only 2.2 kernels that have initiated
this thread...
--
73 de Robin. G8ECJ Hub manager gb7ipd
NTS: G8ECJ@GB7TVG.#42.GBR.EU AmprNet: [EMAIL PROTECTED]
Internet: [EMAIL PROTECTED] http://www.gb7ipd.freeserve.co.uk/
Shack: (+44) 1628 533311 Fax: (+44) 1628 850165
Club pages (g4xyw modem etc) at http://www.tvipug.freeserve.co.uk
