Discussion in 'Linux Networking' started by Chris Rennert, May 10, 2005.

  1. Why do you think so?
    Yes; you cannot use the "/xx" - Notation with discontinuous netmasks.
    Of course. But that does not prohibit using them...

    But nevertheless that 'discontin{u}ous netmask' ist (was) allowed, IIRC.
    A short visit ti google shows a lot of entrys:

    Of course, some hard/software will fail:
    {search for. "discontinous netmask causes PDM to be stuck in a loop"

    Somewhere I've read, that the US Army used such beasts in their net-

    Yours, Holger (dg3lp)
    Holger Petersen, May 12, 2005
    1. Advertisements

  2. Chris Rennert

    Guest Guest

    Say what !

    There is no such thing as an invalid netmask.
    Shows You don't know anything about the subject
    Guest, May 12, 2005
    1. Advertisements

  3. Chris Rennert

    Guest Guest

    that was not a response to Your problem but a reply to Wolf's stupid
    comment !

    I can see You've got quite an attitude when asking for FREE help
    Hire a consultant -- What a lame company
    Guest, May 12, 2005
  4. Chris Rennert

    Guest Guest

    Guest, May 12, 2005
  5. Chris Rennert

    James Knott Guest

    There was absolutely nothing in your original post, about "proof of
    concept". You were describing a situation where you were running out of
    addresses and coming up with what is not an appropriate solution for your
    needs. How are we to know what you want, if you don't express your
    question correctly?
    James Knott, May 12, 2005
  6. Chris Rennert

    Guest Guest

    WHere Did You get This Idea from ????
    Guest, May 12, 2005
  7. You are right, and I do apologize. I should have been more clear in my
    original post. I truely do apologize and appreciate all the help I have

    Chris Rennert, May 12, 2005
  8. Splitting off separate physical networks is almost certainly
    the right thing to do. Despite the fact that you *can* put
    everything on one physical network, it is not necessarily a
    good idea just because network masking allows it.

    If you have a network where the number of IP addresses is too
    great for one 24 bit network mask, it is almost certain that
    other factors besides the simple address count are *more*
    important when laying out the topology.

    For security you probably want 2 or more physical networks. For
    maintenance, for traffic management, and for future additions,
    you'll almost certainly want 2 or more physical networks.

    Don't think about how to go backwards with netmasking, but
    rather how to even diversify further.
    Floyd L. Davidson, May 12, 2005
  9. Chris Rennert

    Wolf Guest

    that was not a response to Your problem but a reply to Wolf's stupid
    I think changing to an RFC1918 B is the right solution and what I would
    recommend and what I would do. You have to touch all the boxes anyway (even
    if only for a program) so why not. Failing that, I would use
    since that would fix the .1.0 and .2.0 parameters and still leave room to
    play with. Successful subnetting has a lot of rules and seem well versed in
    them. Try using a subnet calculator to help:
    Wolf, May 13, 2005
  10. Chris Rennert

    Moe Trin Guest

    Actually, it is valid as far as RFCs are concerned. For your reading
    enjoyment, see RFC0917, RFC0932, RFC0936, and RFC0940, which were the
    conceptual documents, and then read RFC0950 which is the current
    authoritative standard.

    0950 Internet Standard Subnetting Procedure. J.C. Mogul, J. Postel.
    Aug-01-1985. (Format: TXT=37985 bytes) (Updates RFC0792) (Also
    STD0005) (Status: STANDARD)

    specifically relating to the diagram on page 5 of that document. From the
    glossary in the back (because it's easier to read)

    Subnet Field

    The bit field in an Internet address denoting the subnet number.
    The bits making up this field are not necessarily contiguous in
    the address.

    and also RFC1122

    1122 Requirements for Internet Hosts - Communication Layers. R.
    Braden, Ed.. October 1989. (Format: TXT=295992 bytes) (Updated by
    RFC1349) (Also STD0003) (Status: STANDARD)

    specifically, the last sentence of the third paragraph of section

    This notation is not intended to imply that the 1-bits in an
    address mask need be contiguous.

    Notice that RFC1122 is also defined as a standard, and is current. Now, the
    "real world" is described in RFC1219

    1219 On the assignment of subnet numbers. P.F. Tsuchiya. Apr-01-1991.
    (Format: TXT=30609 bytes) (Status: INFORMATIONAL)

    which does indeed _recommend_ a contiguous mask (which is what any
    sane O/S now expects), BUT notice that this document is rated as
    "INFORMATIONAL", not a "STANDARD" as are the others.

    Bottom line - you can use a non-contiguous mask on your network if you
    please (though it won't be easy to configure), because your internal
    mask values are not seen on the Internet. Your network, your configuration,
    your problems.

    Old guy
    Moe Trin, May 13, 2005
  11. Chris Rennert

    Wolf Guest

    That should have been "some of you do not seem well versed in them." *Very
    loong day*
    Wolf, May 13, 2005
  12. Chris Rennert

    Moe Trin Guest

    Cite? Om the other hand, see my response to Ulf Volmer up-thread for
    the RFCs involved. And yes, I am aware of section 4.1 of RFC1519. However,

    1519 Classless Inter-Domain Routing (CIDR): an Address Assignment and
    Aggregation Strategy. V. Fuller, T. Li, J. Yu, K. Varadhan. September
    1993. (Format: TXT=59998 bytes) (Obsoletes RFC1338) (Status: PROPOSED

    it's a _proposed_ standard, while RFC0950 and 1122 are both standards.
    Read RFC2026 to see the difference.
    While no one in their right mind would really want a non-contiguous mask
    (and all current operating systems expect this), it really is permitted.
    It's just not recommended.

    Old guy
    Moe Trin, May 13, 2005
  13. Chris Rennert

    Unruh Guest

    Well, it also depends on how the routing software has been written on your
    machine. If it was written expecting contiguous addresses, then despite
    what the rfc says, you will probably have problems. If I had a dollar for
    eery program which violated the rfcs I think I would be rich.
    Unruh, May 13, 2005
  14. Chris Rennert

    Tauno Voipio Guest

    It's called CIDR, Classless Internet Domain Routing.

    All the ones in the mask must be on the more-significant
    side of any zeroes. No zeroes are allowed between the ones
    and no ones between the zeroes in the netmask.
    Tauno Voipio, May 13, 2005
  15. Chris Rennert

    Guest Guest

    Well thats not the rfd standard ! Go Read the standard !

    The mask is valid but non-continous .. Not to mention You
    can do anything You want on a private network
    Guest, May 13, 2005
  16. Chris Rennert

    Wolf Guest

    It's called CIDR, Classless Internet Domain Routing.
    On the one hand you mention standards, on the other you say you can do what
    you want. :)

    It is not a valid netmask. .Using CIDR or Classful you can never get to a
    253 netmask If you don't believe me, try doing the math with paper and
    pencil. A little demonstration will go a long way to clearing up

    An IP address is divided into two portions: Network and Host.

    When subnetting, you are moving host bits and moving them to the network
    bits. Bits are moved left to right of the host portion, in order, to the
    network section. So for instance a /24 has 24 bits in the network portion
    and 8 bits in the host portion. To subnet this into two networks, you take
    one bit and move it to the network portion and you now have a /25 and 7 bits
    for hosts. Network 0 has 128 IPs and network 1 has 128 IPs (we can forget
    about network and broadcast addresses for the purpose of this discussion)

    When supernetting you are doing the opposite. Moving network bits to host
    bits. The bits to be moved are taken right to left of the network portion,
    and they must all be taken in order. So the /22 I proposed takes 2 bits or 4
    networks. I had to memorize all this junk so I do it in my head, but you
    can just look it up and see that a /22 is a 252 netmask. The next netmask
    down /23 is 254. So, includes the range -

    I will try and find some decent sites on subnetting and post URLs here for
    ya. You can start by checking out Cisco though.
    Wolf, May 13, 2005
  17. Chris Rennert

    Moe Trin Guest

    I don't know of any O/S at the moment that will tolerate a non-contiguous mask
    Nah, only Bill Gates gets to milk that cow, and mainly because microsoft
    goes out of their way to avoid following standards - even their own. Remember
    the compatibility of MS-CHAP-v1 (chap-80 loosely described in RFC2433) verses
    MS-CHAP-v2 (chap-81 even more loosely described in RFC2759)?

    Old guy
    Moe Trin, May 14, 2005
  18. Chris Rennert

    Moe Trin Guest

    Well, technically, it is correct. You might encounter a problem if you
    use a non-contiguous mask, and no current operating system that I'm
    familiar with wants to see it, but "your network - your rules". Just
    don't expect everyone to fall all over each other trying to talk to
    your network.
    Please cite the STANDARD document that replaces or otherwise obsoletes
    RFC0950 and 1122. Kindly consult the RFC Editor index at, or If
    you haven't been there before, be advised that there are a _lot_ of RFCS.
    When I grabbed a copy of the index about 2 weeks ago, the count was a bit
    over 4100.
    That's great. Now go read RFC0950, RFC1122 and the INFORMATIONAL document
    RFC1219. After that, you can then read the PROPOSED STANDARD documents
    RFC151[789]. Then read RFC2026 (note - I'm not shouting, I'm using the
    capitalization specified by the RFCs).

    You are correct that the 'defacto' standard is a contiguous mask. I agree
    with you. Current computer systems and many routers would barf if they saw
    a non-contiguous mask. But RFC0950 and 1122 are still the standards and they
    do not _REQUIRE_ a contiguous mask - see page 5 of RFC0950, and section of RFC1122.
    1878 Variable Length Subnet Table For IPv4. T. Pummill, B. Manning.
    December 1995. (Format: TXT=19414 bytes) (Obsoletes RFC1860) (Status:

    Find RFC1878 at your favorite mirror.

    Old guy
    Moe Trin, May 14, 2005
  19. Chris Rennert

    Wolf Guest

    You have not been doing your homework. Have you actually tried to get 253
    netmask using "the bits?" You can't get there.

    This should help: Take out a sheet of paper and start by creating 8
    columns. This is going to be the 3rd octet. On the top line we write the
    CIDR Prefix from left to right. Next we enter the bit place value.... Then
    add the mask values..

    /17 /18 /19 /20 /21 /22 /23
    128 64 32 16 8 4 2
    128 192 224 240 248 252 254

    Discontiguous networks are ugly and a sign you did not do your work right.
    That can not be fixed with a netmask real or imaginary. It will require a
    router and a protocol that understands classless routing-that leaves RIPv1
    and IGRP off the list.

    I have a really good book on the subject from a class I took last year:
    "CCNP1: Advanced Routing." It goes into a lot of detail on the subjects of
    VLSM, subnetting, supernetting, route summarization etc. The books for
    CCNA1-4 cover it over and over, ad nauseum-and I recommend them also. The
    actually way I calculate it takes another couple of lines. I had to be
    able to do these by hand without using a calculator-in the 21st Century!
    Today my phone has a subnet calculator.

    Another Old Guy,
    Wolf, May 14, 2005
  20. Chris Rennert

    Guest Guest

    Like I said Go read the rfd standard .. what is allowed is not
    necessaraly what is normal done.
    This is not a discussion of LETS BUY CISCO - I don't care what cisco
    does. If they can not split a class into a non-binary number of networks
    that's a good reason not to buy cisco.

    Bottom line
    The original start of this thread was to add more hosts to an
    existing network. NOT how to properly set up a subnet. Hind sight is
    always 20-20
    The easy way is to just change the netmask NOT assign new IP addresses
    to every machine.

    NITPICKING over netmask number's adds nothing to solve the original
    problem. Which was the Whole point of my original post. Just use a
    different netmask. Bickering about which number is a waste of time.
    Guest, May 14, 2005
    1. Advertisements

Ask a Question

Want to reply to this thread or ask your own question?

You'll need to choose a username for the site, which only take a couple of moments (here). After that, you can post your question and our members will help you out.