🚀 go-pugleaf

        So we choose a subnet  mask of 255.255.255.0 for the  individual
    units  and  for  the  lines  we  would  like  to  use something like
    255.255.255.248.  This  way we  get greater  efficient use  from the
    address space, rather than wasting an entire 256 address range space
    for each leased line.  If we have 120 lines, we could end up wasting
    132.64.1.x-132.64.120.x leaving just about 50% of the address  space
    for real work.  Packing the addresses for the leased lines into  one
    segment of the class B network - such as 132.64.1.x makes sense  and
    should work.  But  for some reason  (perhaps some subnetting  rule),
    the cisco will  to not accept  two *different* subnet  masks for the
    same class B  network.  The cisco  automatically changes the  subnet
    mask for all interfaces with the same class B network address to the
    last subnet mask of the last interface set.
    
    I am sure there must be some config command to bypass this problem.
    Or is everyone wasting oodles of address space on their leased line
    subnets?

While OSPF explicitly allows this (as will RIP-II), cisco's implementation
does not support it until release 9.1. As a field-test site for 9.1, I have
just begun to start exercising this feature - on network 131.119.0.0, we now
have routes with four different subnet masks:

	- 255.255.0.0 for the "default subnet" route
	- 255.255.255.0 for variable-length-subnet "challenged" (a California
	  euphamism for handicapped) systems, i.e. pre-9.1 ciscos as well as
	  for Ethernet subnets
	- 255.255.255.252 for serial links
	- 255.255.255.255 for point-to-point connections (originated by the
	  Proteon OSPF implementation)

We are waiting with baited breath for 9.1 to be released, since variable-length
subnet mask support will *greatly* increase the efficiency of our address space
allocation - for example, instead of using up all space in the address range
131.119.1.xxx through 131.119.134.xxx to support 134 point-to-point links (at
2/254 = <1% efficiency), I'll conceivably be able to squeeze all of these
addresses into the range 131.119.1.0 - 131.119.3.xxx at (2/4 = 50% efficiency;
best you can do, since all-ones and all-zeros are reserved addresses).

	--Vince