Re: Whoa; the 3 network?
On Dec 23 11:53, Randall Pigott wrote: % The Princeton address is the same as the old RCA company division % that did DARPA and ARPA gov't contracting, so that address space % once belonged to RCA "in the beginning". False statement and faulty logic. GE merely changed the address for the GE folks handling their network address to be the office in Princeton. Net 3 has always been GE address space, even before GE repurchased RCA during the 80s. In fact, for many years the Net 3 entry at the SRI NIC pointed to a GE Simulation facility in Florida because someone there was the address allocation stuckee for all of GE. % I have personal experience in a past life doing military DARPA work % with RCA, nearly twenty years ago, long before they formed RCA Astro % and built communications satellites. This address space was given % to RCA for DARPA work *only* way back then or earlier. RCA might have been given address space back then, but it wasn't Net 3. In the early 80s, when RCA was not owned by GE, GE had already been alloated Net 3 by the SRI NIC. I was a GE employee at the time and involved in renumbering some internal networks into Net 3 at the time. % I did a casual sequential-countup scripted "ping -a" on a small slice % of 3.0.0.0, and found almost no working domains within this address % space. [stuff deleted here]
-- End of excerpt from Randall Pigott
Ping turns out to be a remarkably ineffective tool at measuring the utilisation of addresses or accessibility of hosts. Many a host that I've managed was not visible via ping, but was in fact directly on The Internet and reachable via telnet/rlogin/ftp _only from authorised hosts_ which were also on the Internet. Connection attempts from unauthorised hosts were silently dropped. No point in making it easy on the bad guys. This was done within some parts of GE at least as far back as the mid 1980s. It helped to have a 4.2 BSD source license. :-) Ran rja@home.net PS: Apologies for posting content onto the NANOG list...
% I did a casual sequential-countup scripted "ping -a" on a small slice % of 3.0.0.0, and found almost no working domains within this address % space. [stuff deleted here]
-- End of excerpt from Randall Pigott
Ping turns out to be a remarkably ineffective tool at measuring the utilisation of addresses or accessibility of hosts.
Ran rja@home.net
True, but "sh ip bgp" is useful. And with tools like the looking glass, there's no room for excuses like "I'm not on a BGP router" or whatever.
From my upstreams (BBN, UUnet, Sprint), I only see announcements for the following traditional Class A's:
{ 4 6 9 12 13 15 16 17 18 20 24 30 32 33 35 36 38 40 44 53 55 57 62 } = 23 used (I'm assuming that the 3.156.20.0/24 from Telstra is bogus) That leaves (128 - { 127 10 1 2 } - 23) = 101 Class A addresses that are unadvertised. Or, to put it another way, 1,694,498,816 available IP addresses, more or less. While I'm sure there's a legitimate reason for some of this to be allocated and not advertised all the time, it sounds to me like there's a big waste of space here. I may be missing a few more reserved nets and whatnot, but either way you slice it that's a lot of address space. "But wait!" I hear someone cry. "I own one of those Class A's that you don't see advertised, and I may want to eventually connect my web server to the Internet some day!" OK. Fair enough. Why don't you keep x.0.0.0/18, and give the rest back? Or find yourself a NAT box? Or (etc...etc...all been said). This is assuming that you only own one netblock, and aren't using another block for your corporate 'Net gateway. So how do we reclaim the unused address space? I dunno - very carefully? Approach all 101 unused Class A contacts, let them know that they'll have to justify their use of space sometime in the next six months. See if you can find one that completely went out of business. Then make sure that people like us, that may filter out BGP based on block/mask combinations are aware of the new allocation scheme. Give out the newly reclaimed Class A until it's almost gone, then start again. Or has this been done? I seem to remember something on this list a few weeks ago about unused address space (maybe in the beginning of this whole 3.0.0.0 thread). If it has been done, why didn't it work and what needs to be done differently? (And I mean that constructively.) eric
hi ran - let's take this off the group and onto private email, ok? At 08:54 AM 12/29/97 -0800, Ran Atkinson wrote:
False statement and faulty logic. GE merely changed the address for the GE folks handling their network address to be the office in Princeton.
Net 3 has always been GE address space, even before GE repurchased RCA during the 80s. In fact, for many years the Net 3 entry at the SRI NIC pointed to a GE Simulation facility in Florida because someone there was the address allocation stuckee for all of GE.
and where did it point even father back when the address space was defined by the darpa host table, updated by crude non-deterministic scripts?
RCA might have been given address space back then, but it wasn't Net 3. In the early 80s, when RCA was not owned by GE, GE had already been alloated Net 3 by the SRI NIC. I was a GE employee at the time and involved in renumbering some internal networks into Net 3 at the time.
you neglected to include the history of the 3 space *before* it was transferred to ge. or the contract task win by ge and the various task rebadgings and reallocations over the research years. ge lost the follow-on but *kept* net 3. my statements still stand as accurate, to the best of my rapidly-decaying memory :) age is getting less kind every day............ look, the whole point of this orig reply was to set the thread off onto a productive course of discussion on how and why to use public ip numbers more efficiently that having dozens of /8 blocks out there apparently with lame allocation or inefficient use. if that got network operators talking about using nat to provide more efficient enterprise networks and then apply those techniques in relevant ways to our own network architectures, then i was successful. that was the only point. whether it was ge or xyzzy is *not* the point. (be sure to leave column 7 blank on the coding form for a single-line entry - or was that column 6?:) there could be well over 10 to 15 billion ip addresses right now being hogged whose use cannot be defended against the greater need of other public network operators. this is itself a corporate theft-of-service, more white-collar and passive a danger to the growth of the internet than spammers but certainly no less an odious crime. combos of nat between rfc1918 hosts inter-firewall and private numbering within the vpn can provide good service to a ge or other wasteful /8 delegees with enterprise networks, with *only* a few classful b's and still provide a good level of security - maybe even better security in some respects. it is irrelevant to the argument whether one uses ping only because it was not worth more than two minutes of one's time, or if one works more in depth and spends more time with sh ip bgp or looking glass tools. i was not seeking to perform an in-depth quantitative analysis, and most readers knew that. everyone knows few public hosts still have icmp enabled in these new and scary times. the point remains - this issue of wasteful and indefensible ip hoarding by legacy delegees needs spading to get the worms and parasites exposed to the sunlight and fresh air. that the hoarders are large corporations is no excuse, in fact it makes their waste and abuse even less tolerable. that issue *is* relevant, and discussion of network architectures to more efficiently use ip number resources is very much on-topic to nanog. now, let's give randy bush a break and take any more p*ssing about ge vs. rca to private mail, ok? :) nanog subscribers have too much to read without learning the history of gov't contracting 101 on the listserv.................. happy new year, everyone! randall
participants (3)
-
Eric Osborne
-
Randall Pigott
-
rja@corp.home.net