[CLUE-Tech] WAS: Netmask trivia question...

[Keith Hellman]

On Sun, Aug 31, 2003 at 12:37:16PM -0400, black at galaxy.silvren.com wrote:
> Hi, not to start stepping on toes, but netmasks quite positively have to
> be contiguous 1's. 255.255.1.0 and the like are impossible. You can't
> borrow bits for host addresses arbitrarily.
> 
> 255.255.1.0 is a valid HOST address, but it is most certainly not a valid
> netmask.

And I quoth:
  "In theory, IP allows the mask to have noncontiguous 1's.  This means
  that the bits that indicate the link number need not be contiguous but
  can be sprinkled thoughout the address."

  Interconnections (Bridges, Routers, Switches, and Internetworking
  Protocols), 2nd Edition.  Radia Perlman, Addison-Wesley, ISBN
  0-201-63448-1  p. 195

Admittedly, she goes on to point out that it is ill-advised to use
non-contiguous netmasks:

  "...and in fact none of the router vendors supports noncontiguous
  subnet masks, so we really can think of the link address as a prefix."

Radia Perlman, IIRC, is the lady that developed the layer 2 bridging
algorithm, and is still actively involved in network research.  I highly
recommend the book BTW.

-- 
Keith Hellman                             #include <disclaimer.h>
khellman at mcprogramming.com                from disclaimer import standard

"You want to know the secret to immortality, write a bunch of songs that people
keep singin' and playin'."

-- Dr(x) John on Duke Ellington