🚀 go-pugleaf

In article <1992Sep2.081216.22430@nuscc.nus.sg> ccewch@nuscc.nus.sg (Wong Chee Heng) writes:
   When a PC dial into a unix host using telnet over PPP, the PC
   appear as a IP node right? What about the UNIX host? It becomes a
   IP router right?  If that is the case then each modem connected to
   the UNIX host will be having a different network interface ...

Right and Right and "Right so far".

   which need a different subnet number!  is it true?

Not necessarily.

   Or all the modems/dial-in lines can be considered within one
   network separated (different from the subnet number the UNIX host
   is in) and share one common IP subnet number, 

That's one way to do it, and it fits well with the normal way IP
routing is handled.

   or in fact all the dial-in PC can share the same subnet number of
   the UNIX host and become part of the same network the UNIX host is
   in?

That's the other way to do it, and it fits well enough with normal IP
routing by faking direct connectivity to the office network.  Just
instruct the office UNIX host to return its Ethernet address in
response to ARP requests for the PC's IP address (e.g. `arp -s
pc-IP-address unix-ether-address pub') so that everything else on the
office network will know how to get to the PC.

One big advantage of the ARP-but-don't-route technique is that it
requires no change to your routing tables, since you're not really
routing IP packets between networks.  Another big advantage is that
you needn't apply to your campus network administrator for Yet Another
Subnet for your department's own use, particularly since a
dialup-based subnet's address space will tend to be even more sparsely
populated than that of a subnet on a LAN, which may cause the network
administrator to be reluctant to hand you another subnet number.

In our network, do it both ways.  Some of our home Suns are on the
same IP subnet number as the main office backbone Ethernet, and some
are on their own subnet number.  They all dial into the same modem
pool, attached to a Sun.  For more discussion, you might find some
useful tidbits in section 7.1 of the Morning Star PPP User Guide,
available from ftp.morningstar.com:pub/ppp/user-guide-2up.ps.Z.

   Whichever case is it, how does it affect the IP routing table? Is
   static route required, or RIP is used?

Our in-office network is simple enough, and the routes stable enough,
that we use static routing.  Some users run routed on their UNIX hosts
and talk RIP with their routers.  Users with even more complex network
topologies, and with even greater requirements that the UNIX hosts
participate in the routing infrastructure, prefer to run gated and
talk any of gated's myriad routing protocols, including OSPF.  Use
whichever you need to solve your particular problem.  If the route is
in the kernel's routing table, IP doesn't care how it got there.