[KLUG Members] some network questions

[Adam Tauno Williams]

> I have a question that I am puzzeled with for some time now. I hope
> somebody on this list has major knowledge on this subject.
> The network/host part is set by the netmask, what basically means that the
> last bit set (from left to right) marks the end of the network part. What
> Is this behavior different along different operating systems? 

No, it should be the same.

> It appears that routers can't see hosts on an other site if that host has it netmask
> wrong.

Of course.

> A second question related with the first, if a number of sites are linked
> together via routers, the connection fails when compression is used on the
> lines. -unless- rarp is switched on. I can only explain this behavior if
> switching compression on means that the networkpart bits are left out in
> the transmission. Is this true?

No, you have something miss configured.  The protocol of the link cannot
(as in 'is forbidden to') change the content of the stream.  You are not
reassembling fragments, not agreeing on the MTU/MRU, or you have a
netmask or subnet mismatch.   ARP, or RARP, shouldn't have anything to
do with a WAN in an IP network.  If you need to fiddle with RARP at all
then you have something wrong.

> Next to the fact that not sending bytes is the most effective way of
> accomplishing less trafic, the routers only need to exchange there network
> parts on initialisation.

?

> With rarp switched on both sites are using tables to hold the mac/ip
> translation as I am told. 

Yep.  But why?  This is pointless.

> But than using rarp should mean even less trafic
> is needed, the whole ip-info can be droped on transmission.

No, it doesn't work that way.  The information in the header of the
packet contains lots of required information.

> But that is not the behavior observed. Switching rarp on increasses the
> trafic dramatically.

Of course.  You aren't running a routed network but instead a big
spudded to together LAN like PC networks circa 1991-1992.  Just
configure your network in the correct current-practices way of a routed
IP network.  There is ******NO******* magic bullet to make a circuit
carry more data than it will carry;  buy any black-box from anyone and a
128k circuit is still a 128k circuit.

>  ISDN lines are almost always open instead of closed.
> What does make the router decided to open up the lines?

Traffic.  Which, in an enormous broadcast domain like you've created
with RARP and what-not, will be constant.  Also if you have windows
machines on those demand circuits you need to change the NetBIOS node
type to disable broadcast browsing.

> So rarp could decrease the needed bandwith at any time but increases it at
> other moments; eg some sort of leveling of traffic.

You don't need it.

> Also, hosts having there netmask wrong are seen from the other site.
> That's nice but following the advice of a telco engineer just to always
> use rarp "because it than always works" (he said there standard procedure
> was always using rarp), does make a nice bill, but does not make me happy.

You are taking advice from a telco engineer?!  He says that because they
don't know how to configure a routed network.  Never let a telco person
within 15 feet of your router.  If need be just install an automated
motion-activated machine gun turret and mow them critters down into
crimson sticky paste.  I'll never forget the day I looked over the
shoulder of an SBC tech at one of our locations who way keying in static
routes to all our locations... this was the next step on the check list
AFTER enabling OSPF routing. 
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 189 bytes
Desc: This is a digitally signed message part
Url : http://www.kalamazoolinux.org/pipermail/members/attachments/20051122/aa1a94d7/attachment.bin