Re: Policy Statement on Address Space Allocations
From: Geoff Huston <G.Huston@aarnet.edu.au> it will be economically infeasible IF we continue with this strange system of zero dollar interconnections we use as a peering model. ... In the same way that giving away IP addresses and giving away IP routing can only be described as a very bad case of irrational behaviour, especially when the underlying resource is under stress ... then I'd also note that giving away transit is similarly a case completely irrational behaviour! All this points to a desperate need for a more realistic economic structure to be used within a number of key aspects of Internet infrastructure. Geoff, while I sort of agree with you, please note that the current "free" models in all these areas do have one real advantage, which is that they are simple. Perhaps we will need "more realistic economic structure[s]", but I can more or less guarantee you that those will be more complex as well. Also, that complexity has a cost, which also has to be figured in. (The classic example is long-distance telephone service, where the cost of creting the bill is said to be far larger than the cost of the underlying service.) Finally, given past history, I suspect that the 'Net community, and particularly the 'Net technical community, is going to have a painful transition, on the emotional level and others, to this brave new world of more commercial structures, and the inevitable technical impacts that causes. That has certainly been the case in every such step in the past... Noel
All this discussion seems to be about work arounds for the real problem. Namely, that the current hardware/software/protocols can't handle what is actually a small number of routes. Restricting announcements of new routes should be one of the last things considered. Here is one - for whatever reasons, we have a provider who can't seem to correctly announce an aggregate and instead, announces several specifics. Nobody says anything about this inefficiency - not to us or to them. Some automated process watching for things like this and sending an advisory email might help quite a bit. -- (313) 741-4442 http://branch.com/ Jon Zeeff Branch Internet Services Inc. jon@branch.com *** WWW Hosting Services, WWW Site Development and the Branch Malls ***
In message <m0thjaK-000NizC@aero.branch.com>, Jon Zeeff writes:
Absolutely. The problem stems from inadequate foresight on the part of router vendors and providers being unable to sufficiently influence router designs so that the needs of high end providers are met. There isn't a whole lot of viable choices at the high end.
Restricting announcements of new routes should be one of the last things considered.
I fully agree. It was one of the last things considered. Quite a while ago on this list it was pointed out that address leasing and coerced renumbering (its coming down to forced) was something we wanted to prepare the community for but that we were hoping to avoid. It might be that better routers and/or better methods of configuring aggregation help take some of the pressure off and change things back from "forced renumbering" to "encouraged renumbering". That hasn't happenned yet.
Tony Bates used to do this and send it to the list. Its not as if no one has thought of this. Curtis
To the list? I don't see what good that would do - it should go to the owner of the addresses and the parties announcing it. This hasn't been done. Another factoid to consider - I know of a company that has a Class C that they don't use. To my knowledge, nobody has ever even asked that they give it up. Some automated email process could do this without much effort. I (and others) have suggested forced aggregation except for routes that are specifically registered as "don't aggregate". I haven't seen any movement towards that.
Worse yet, I applied for several class-C's for companies years ago (before the CIDR stuff really got off the ground) that I _know_ are obsolete, but the Internic won't let me recycle them. I should have left my name on them in some way I guess. The problem is, the address listed is obsolete, the phones disconnected and in one case the contact is dead! So, how is one supposed to try and recycle in this sort of situation? On Thu, 1 Feb 1996, Jon Zeeff wrote:
Ed Morin Northwest Nexus Inc. (206) 455-3505 (voice) Professional Internet Services edm@nwnexus.WA.COM
Well, almost. The IPGR robot is, in fact, doing just that. As an aside, for my own private entertainment, I'm conducting a straw poll. If you beleive that prefix filtering is the right way to address this problem: or If you beleive that prefix charging is the right way to address this problem: or Gentle Renumbering requests will work. Your selection to me in private mail. Door prizes at NANOG for the attending participants. -- --bill
"All in good time my pretty..." We are working on the 192.x.x.x swamp right now. Rough estimates (with much more accurate data @ NANOG) 60% - invalid or missing contact information 25% - in use & unwilling to renumber 15% - willing to renumber or return This is from ~6,000 delegated entries. If all goes well, we can find some new area to work on sometime in later this year. Have any suggestions? 198.x.x.x? The old /16 space? --bill
Interesting indeed. Lets see... Nobody can contact you .. is that the admin/tech contact, the administrative entity (corp, gov, agency etc) or ???? Addresses not reachable .. From which vantage point is this measuerment taken? Some period .. Like the 99 year lease on HongKong? Perhaps there is better wisdom out there on correct metrics for these values. From my limited viewpoint, the only way to recover the space is a voluntary return, based on the original allocation policies. There may be other incentives applied to facilitate the return, but strong-arm tactics and coersion, threats and hostile actions are not my favorites. I'd prefer to take almost any other action than blacklisting and hijacking. To take such actions, while it can be rationalized as a technological means to protect a networks internal stability, is presumptious and rude at best and legally indefensable at worst. Now if there are existant policies -in place-, that constrain the prefix handling, then your questions have been answered. Just my humble opinion. --bill
There must be some mechanism implemented whereby address space will return to the IANA after a specified period of time unless otherwise requested by the prefix holder. Otherwise what will happen is that this 60% (or some other large percentage when the figures finally settle) of the 192/8 address space will effectively be lost from the internet with no real means of retrieving it. A system like this without any garbage collection mechanism is eventually going to fill up with defunct allocations and the cruft of years past -- something which is not an option when dealing with limited address space.
How are the InterNIC coping with the new domain name charging scheme? If this were successful, a similar scheme might be considered for address prefixes. The legal consequences are similar if not quite the same, and one is really no more rude or presumptious than the other. Nick
There were a couple of methods suggested here: preemptive hijacking - voluntary return - periodic fees - Hijacking has a number of interesting problems Periodic fees will take a year or more to implement Voluntary return can be done -now-. Which method is the least stressfull and has reasonable impact on the existing routing table crunch?
Not quite the same beast. Domain lables are -not- a finite resource. There are a wide range of viable alternatives to paying the InterNIC fees.
Nick
-- --bill
That depends completely on whether you're firefighting the current urgent problem, or else taking steps to ensure that it's not going to happen in future. Ideally both courses of action should be taken. We're simply discussing what the long term approach should be. Hijacking is not an option. This would *really* mean the end of the Internet as we know it, and there really would be news at 11! :-) Voluntary return is a really good idea, but it relies on good-will (and good manners) to get a return. And I'm quite surprised that there were as many as 15% of prefixes returned in such a short space of time. But periodic fees or some similar scheme is necessary. The matter really boils down to two different psychologies: a) The IANA IP number allocations stay in effect until otherwise notified, or b) The IANA IP number allocations only stay in effect as long as the IANA are kept informed that they are still in use. Voluntary return assumes a), and periodic fees (or similar) assumes b). a) will get a certain amount of response in the short term, and after that, the response will be quite low. b), on the other hand, will gain a very high response in the short term, and as-high-as-is-possible reponse in the long term, albeit at the cost of setting up the administrivia to ensure that the scheme worked. It's a much more aggressive approach, and it may upset some people initially, but long term, it could well turn out to be the only viable option.
There are a wide range of viable alternatives to paying the InterNIC fees.
Perhaps, but any system which might be implemented to return IP allocations will take time, effort and hence money. That the InterNIC should be paid for their efforts is not unreasonable. Nick
On Fri, 2 Feb 1996, Bill Manning wrote:
You don't neccessarily need addresses returned to solve routing table crunch. If contact with the address owner really is impossible, then they are not using the addresses on the global Internet and therefore their addresses can be aggregated with other live addresses. The best way to do this would be to move all live addresses out of the block and just drop the whole block for the time being. Of course this means that other people would need to renumber into more router-friendly prefixes but that's what PIER is all about. Michael Dillon Voice: +1-604-546-8022 Memra Software Inc. Fax: +1-604-546-3049 http://www.memra.com E-mail: michael@memra.com
In message <199602021407.AA25614@zephyr.isi.edu>, Bill Manning writes:
Bill, There is no need to call it hijacking. If an organization registered an address they are responsible for keeping the contact name up to date. If they don't announce the route, they have not provided a valid contact, and there is no way to contact them, including publishing a list on major mailing lists, then it should be safe to recover the address since every reasonable effort was made to contact them. If a route is not announced, this is a NOOP anyway. If the route is announced, go through the AS path and/or traceroute asking the provider closest to the route for a contact name. Just send a "Dear IP Provider" letter stating "This appears to be your customer but we have no way to contact them. Can you help?". Most providers have a way of contacting their customers. This should help with the 60% that can't be contacted. Yes - I know this is work, so don't take this as a complaint that you are doing something you should be, just a suggestion for dealing with this problem. Curtis
The kicker is, where are they keeping the data? InterNIC ? DDNnic? RIPEncc? The problem is compounded with the InterNIC and the DDNnic keeping authoritative data over the same space. Can you say "SRI connected/unconnected database problems"... sure you can.
In fact, that is exactly why a robot mailer is not a cureall. The process followed is close to yoru description. Hence a slower pace of progress than many would like. This swamp is -deep-. --bill
Curtis You ignorant bozo - I meant it has no effect on routing table size. Any attempt to recover it is a NOOP until we actually need the 192/8 space which won't be for a very long time (after A space is exhausted). ;-) -- Curtis ps - I trust its perfectly acceptable to email flaming insults when replying to yourself.
On Thu, 1 Feb 1996, Bill Manning wrote:
so what you're saying is, if a Government (agency) were to take such action, it could work..??? but then again, we don't *want* government involved... where's daddy when you need him ? Love/hate relationship to say the least.
On Wed, 14 Feb 1996, Edward Henigin wrote:
well, then there are a couple of simple rules to be observed 1) dont piss off the international community and declare Internet to US owned (like the guy who posted 'he is sick of his country to provide support for others who don't deserve it'. 2) work for seamless interoperability, and this means cast out people or corporations who do noe 3) don't implement anything that causes friction , is not backwards compatible, or deprives communities (which can be nations, networks, religious aggregations or sex maniacs) from expressing themselves *between themselves* (watch the stars!). Be conservative in what you receive, leading edge in what you provide Simple, no? Mike (this .sig is really only for id, not for anything else!)
where's daddy when you need him ? Love/hate relationship to say the least.
---------------------------------------------------------- IDT Michael F. Nittmann --------- Senior Network Architect \ / (201) 928 4456 ------- (201) 928 1888 FAX \ / mn@tremere.ios.com --- V IOS
How about a policy which says that if you fail to pay an annual amount by a due date then a process is commenced with an outcome such that your registration details expire and the associated number space is no longer registered to you. Geoff
By addresses not being reachable, are you effectively saying that any enterprise that does not want to connect to the Internet must use RFC1597 address space? Anyone have an idea how much of the address space is used for registered addresses of organizations that do not connect to the Internet? This is not a trivial question, because I am aware, at least, of an assortment of military networks who have registered addresses, connect with other arbitrary military networks with their own registered addresses, and really need some assurance that these internetworks will have unique addresses. Internetworks != Internet, so valid assignments may not be Internet reachable.
On Fri, 2 Feb 1996, Howard Berkowitz wrote:
I would also be curious how the 60% missing is counted. If an organization places 99% of their addresses behind a firewall do all those not count? Unfortunately, I don't think we can base much policy on whether or what % of addresses are reachable from the internet. --- David Miller ---------------------------------------------------------------------------- It's *amazing* what one can accomplish when one doesn't know what one can't do!
On Thu, 1 Feb 1996, Jon Zeeff wrote:
Ah, but yes it is... The other day, I received the following message below asking us if we were still using the network provided to us. Since we are using it, we politely replied; I'm sure many more people than I were contacted for older networks. /cah ---- Craig A. Huegen <chuegen@pyramid.com> Phone: (408) 428-8404 Communications Engineer Fax: (408) 428-8513 Electronic Data Systems / Pyramid Technology Corporation Mail Stop SJ1-1-107, 3860 North First Street, San Jose, CA 95134 --- Begin enclosed mail --- Date: Mon, 22 Jan 1996 11:46:37 -0800 From: ipgr@ISI.EDU To: chuegen@pyramid.com, ipgr@ISI.EDU Subject: Network Number Usage Survey-- 192.107.50.0 Hi, We have been asked by members of the PIER Working Group of the IETF, with the approval of the IANA and Internic, to conduct a survey of a section of the IPv4 address space. Your address appeared in the InterNIC database as the likely person to ask about the following set of network numbers: Pyramid Technology Corporation (NET-RELIANT-HV) Network number: 192.107.50.0 If you are not the correct contact, please forward this message to the right person if you can. If you are, we would like to know: Is your organization still using this address space? If you are not using it-- would you be willing to return this address to the IANA for reallocation? Your answers are important in planning future allocation of the IP address space. Thank you for your time. --- End Mail ---
participants (14)
-
bmanning＠ISI.EDU -
Craig A. Huegen -
Curtis Villamizar -
David Miller -
Ed Morin -
Edward Henigin -
Geoff Huston -
Howard Berkowitz -
Iljitsch van Beijnum -
jnc＠ginger.lcs.mit.edu -
jon＠branch.com -
Michael Dillon -
mike -
Nick Hilliard