Netmasks for dummies 3012

The one that basically says 'yea' or 'nay'' is 0950, and the listing from the current index says:

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

It's ancient, but it's current. But then, so is 0791 (Internet Protocol) and 0793 (TCP) which both date from September 1981. If you are interested in the background of the network mask problem, grab copies of RFC0917, 0925, 0932, 0936 and 0940 (Stanford, ISI, MIT, Berkeley, and SRI concepts respectively). 0925 is roughly proxyarp, while the others use the routing tables. None mention the consecutive ones concept, and in fact seem to want to use an octet to give 127 or 256 subnets. Briefly, this would be no different than an existing routing table, except that there would be lots of entries, one for each subnet (or a single default, although this isn't really spelled out clearly). I suspect this is why it's not defined in RFC0950. If you have a copy of 'TCP-IP Illustrated Volume 1' by W. Richard Stevens, he even used this as one of his chapter questions (3.5 in my 1994 printing). His answer: legal but frowned upon.

I didn't fall into this network) end of the morbutt until about '93, and the lead people I worked with had already installed some interesting firewall rules which basically prevented nearly all incoming connections to the8. What limited access we did get from outside was a bastion host with a SecureID login (obviously not many had that access). This was a good idea, because behind that firewall we still used NIS, and the Berkeley 'r' commands were available, even as root. While the mail and news headers had full hostname and IP (till someone pointed it out), those names and addresses didn't even resolve from outside, never mind being reachable.

While the people who created the Internet (DARPANET, ARPANET, whatever you want to call it back then) were brilliant, the climate they worked in was also different. Some people had rather 'strong' opinions - some one way, some the other. There's plenty of indications of this in the various documents, so I don't have to kick any of the ghosts in the closets. I suppose the results were the best we can hope for. Still, reading RFC0089 does bring a chuckle to some.

Isn't it. RIPE seems to be willing to allocate these weird blocks, like the above, and '96', '140', '192', and '240' hosts. At one time, there were several blocks of '27' allocated, though I forget where.

I'm not even sure it's possible with any operating system today. None of the commercial Unix I have access to will, and this is also true for Linux. You can use logical 'AND' and 'NOR' to read a mask and IP, but I can't imagine someone send enough to write a network stack not using the consecutive ones method. It makes routing decisions so much easier.

I was just trying to answer Bill's question about it being legal to have a non-consecutive mask. While RFCs do not have the power of 'law', they don't absolutely prohibit this - they just make it reasonably obvious that it's not a "Good Idea"(tm). The use of non-consecutive ones in the mask would be more complex, but Stanford and Berkeley seemed to have it working back then.

Read RFC0936, and then ask what they were thinking about.

Old guy