February 2001 - lir-wg - mailman.ripe.net

Summary: Fixed Boundary (/29) Assignments
by Mirjam Kuehne 19 Sep '01

19 Sep '01

Dear LIR-WG, Here is an attempt to summarise the recent discussions on this mailing list regarding possible fixed size assignments. Following a review of the comments, it looks like the majority does not believe that a special assignment policy is needed. Most contributors agreed that there is no need to make a distinction in policy relating to different access methods. All assignments should be made on actual needs and according to the current IP address policies. The comments showed concerns about the administrative procedures required for these kinds of assignments. However, it was felt that these addresses need to be accounted for in some way. A possible way forward would be to initiate discussion surrounding the set of information necessary to justify these assignments while providing an efficient way for the LIRs to provide services to their customers. I realise that I did not touch on every aspect of the discussion but I hope that this is a fair summary in order for you to move forward Looking forward to a continuation of this debate. Best Regards, Mirjam Kuehne RIPE NCC

3 2

Re: Modifying references to contact info in the RIPE DB
by Wilfried Woeber, UniVie/ACOnet 24 Jun '01

24 Jun '01

Hi Joao! >If the reference is by name then when I query the DB I will get the list of >persons with that name, right? Right. >Whether the object belongs to the person or not is not going to save this >person from being contacted. From the point of view of the database and the >user querying it, the object belongs to that person. Right >The link is via a name. > >Now we substitute the reference by name with a reference by nic handle >pointing to the same person. >What is the difference? > >ALMOST NONE except that: True, for the case where there is a strict one-to-one mapping, e.g the recursion using the name returns *one* object. What do you suggets to do with the case where the recursion, based on the names, returns a *set* of objects? I can see 2 approaches (at least): - do not perform the handle substitution for that referring object and put it on the list of those requiring human intervention - expand the xxxx-c: list (which? the tech-c:?) to refer to *all* objects, but by ref'ing their handles While human beings are quite clever in selecting the "proper" instance of a person: object (by looking at country, address, email,...) a script might be quite clueless with that respect :-) Wilfried. -------------------------------------------------------------------------- Wilfried Woeber : e-mail: Woeber(a)CC.UniVie.ac.at Computer Center - ACOnet : Tel: +43 1 4277 - 140 33 Vienna University : Fax: +43 1 4277 - 9 140 Universitaetsstrasse 7 : RIPE-DB (&NIC) Handle: WW144 A-1010 Vienna, Austria, Europe : PGP public key ID 0xF0ACB369 --------------------------------------------------------------------------

3 3

Management of the working group
by Hans Petter Holen 04 Apr '01

04 Apr '01

Dear lir-wg, as just expressed at the wg meeting, I think it is appropriate to suggest some formalities for the wg to elect its chairs. My proposal is as following: 1) Open nominations on the lir-wg mailinglist 2) Optional discussion and support statements 3) Election of chair and 1-2 co chairs at the next lir-wg meeting in Bologna. 3b) The election to be performed either open ballot or by secret ballot if demanded by one or more participants at the meeting 3c) The election to be confirmed by consensus by the wg after the election result have been duly counted by a counting team from the audience. 4) That we perform such an election annualy to ensure the community support to the wg chairs. As the meeting expressed support towards such an idea, I would like to hear the opinions of the wg on theese matters before proceeding. Sincerely, Hans Petter Holen lir-wg Chair

6 10

Re: IP Management Tool - Minimum Requirements
by Guy Vegoda 06 Mar '01

06 Mar '01

> Would there not be intellectual property difficulties if you > wrote a similar system for another company? > > Niall > > (Interested but also enmeshed in IP problems) I guess that depends on how similar you mean by similar (-: I don't think IPMT will be anything like Robo-Bijal, code-wise. (Not least because: (1) I will not be the only coder, (2) Manuel's IP libraries are open source, (3) It will probably support IPv6, which Robo-Bijal doesn't, and (4) I would like to do it in OO perl. ) So should hopefully all work out. Guy -- Guy Vegoda \ guy(a)vegoda.org *Please do not send html* NIC: GUY-RIPE \ guy(a)cryptography.org.uk *attachments* Unix, Linux Hobbyist \ +44(0)20 7961 8318 (work) www.thenakedfrenchman.com \ +44(0)958 469 532 (cell)

2 1

Re: Refuse een assignment because it 'cannot' be routed?
by Wilfried Woeber, UniVie/ACOnet 05 Mar '01

05 Mar '01

To continue that (cute | silly | pick-your-choice) analogy... => I only want a fixed frontdoor (1 fixed IP address), but I am trying to force => my landlord into letting me have it without paying tripple rent, by asking => the government (RIPE) to give me a building permission to install more => doors. Not because I want more doors, but to keep the landlord from moving => my single door every day :) => = =Hi, = = Hmm... this analogy isn't correct. RIPE is not the government in this. =RIPE, or your local LIR can give you a door (address assignment), but =you still need to get a permit from the local counsil to place it =(Getting your service provider to actually route it). = =One of the things your local LIR may require before selling you a door =is having a permit. Buying the door somewhere else (RIPE) does not =automagically entitle you to a permit. ...what we are seeing in reality, though, is something like the "government" (the RIPE NCC) coming back with a question about the colour and style of the door. And - depending on your answer - either going ahead issuing a _permanent_ permit for e.g. an Ethernet- or leased-line-style door ("static"), but limiting the validity of the permit for an xDSL- style or dial-up door to as long as you, or someone from your family, happens stay at home ("dynamic"). As soon as you go to work, or even worse - take a couple of days off, duhh!!! - you have to submit another application for installing a door. And, the "government", suggests to the manufaturers of the doors (and/or to the landlord), to give you a call on the phone every, say, 8 hours, to confirm that you haven't left for shopping (physicly) or turned to RTFM (virtually/mentally logging off). That's exactly what happens to me, back home, with my ADSL link being dropped every 8 hours by the ISP *on purpose*, because the NCC sort of "suggests" to the ISPs to use dial-up ratio mechanisms for 24x7 xDSL, flat rate billing. Very clever, indeed, in particular when someone tries to do stuff that is security-aware. But I have beaten that one to death before (technology independence of assignment rules). Sigh... Wilfried. _________________________________:_____________________________________ Wilfried Woeber : e-mail: Woeber(a)CC.UniVie.ac.at UniVie Computer Center - ACOnet : Tel: +43 1 4277 - 140 33 Universitaetsstrasse 7 : Fax: +43 1 4277 - 9 140 A-1010 Vienna, Austria, Europe : RIPE-DB: WW144, PGP keyID 0xF0ACB369 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~:~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ Träume darf man nicht verbieten, sonst werden sie Wirklichkeit...

3 3

Refuse een assignment because it 'cannot' be routed?
by Herbert Baerten 02 Mar '01

02 Mar '01

Dear colleagues, in the discussion about assigning fixed address space to residential always-on customers, I concluded that most ISPs are eager to provide their customers with this service, even using it as a weapon in the competition battle. However in some countries apparently the contrary is going on. All ISPs there offer 2 classes of DSL service: cheap subscriptions (low traffic, dynamic ip, no options) on one hand and an expensive 'pro' contract (high traffic volume, static ip, router etc.) on the other. Now a customer would like to get a static ip but does not want the router, large volume etc so he does not want to pay 4 times as much per month (real-life example) just to get a static ip. If the customers sends a RIPE-141 request to the ISP, can the ISP assign a range but not route it to the customer? Can it refuse the request on the grounds that it cannot route the assigned range to a dynamic ip? Or is the ISP obliged to assign a range and route it (no matter how, presumably by assigning the customer a static ip anyway) ? With or without additional (reasonable?) cost to the customer? tia best regards, Herbert -- HB5351 Herbert Baerten Network Manager HostIT Benelux

4 7

LIR Charging Publications
by Roberto Pezzile 28 Feb '01

28 Feb '01

Hello to all colleagues, I am Roberto Pezzile, network and telecommunications manager of Seven, in Castelfranco Veneto, Italy. We have recently become a new LIR. In the documentation given, I have a few questions regarding the publication of "Charging by Local Internet Registries" - precisely, ripe-152.txt; where "5. Recommendations" urges all registries to publish operating procedures, service details,and relative charges. In my working experience, mostly in Italy, but occasionally in the US, Brasil, Germany, France and China, I have never seen such Charging Information published at all, by anyone. Perhaps I've missed something somewhere. Even reference documents ripe-104 and ripe-112 did not offer additional information . I would like to know: 1. Is there a form to fill out and send to RIPE for eventual publication ? 2. Is there another LIR in our area which can show us an example of their charging policies and procedures ? 3. What is intended by ".....achieving consensus within the community served regarding the disposal of any surplus revenues"? I look forward to receiving information on the above, based on the practical working experiences of those kind enough to give me a hand. Thankyou very much Roberto

1 0

Re: Refuse een assignment because it 'cannot' be routed?
by Wilfried Woeber, UniVie/ACOnet 27 Feb '01

27 Feb '01

> > >> Now a customer would like to get a static ip but does not want > > ONE static IP address? PA, I assume? > > Yes, PA. > > > > >> If the customers sends a RIPE-141 request to the ISP, can the > > ISP assign a > > >> range but not route it to the customer? I guess, formally, the answer is YES. Because when the customer can document his/her need, then the LIR *may* assign addresses But, holding a block of addresses does not in any way "guarantee" connectivity... > > > > ... which would be extremely useful for ??? routing can be very different things: - distribute the announcement for that address space only within the ISPs IGP - distribute the announcement to direct EGP peers, but with "no-propagate", i.e. for providing access to a particaular server, or to suppoort dual-homing/mutual backup - make the address block (potentially) visible from "everywhere", by announcing the encopmpassing aggregate and/or the particular (more specific) prefix. The usefulness is subject to effects of aggregation and filtering - ... > In order to force the customer to buy a more expensive service which > includes fixed ip! Formally, definition of prodcut packages and charging for services is outside the scope of the rules for address assignment - with the exception that the address space itself must not be sold. > Please understand that nor I nor my employer is doing this or even > advocating this practice. To the contrary: I'm just observing that it > happens and am so displeased about this that I am looking for ways to > counter it. As I said in a different thread, this should be discussed in the frame-work of a BCP, either within RIPE or somewhere else. We might also want to leave it to competition to eventually deal with that (don't hold your breath ;-) > > >> Can it refuse the request on the grounds that it cannot route > > the assigned > > >> range to a dynamic ip? > > > > ... so, you want to rent a house and put a security guard on the entrance > > door not allowing people who bought it to move in. Interesting ... > > I think it is more appropriate to state that someone is looking for a house > but all landlords he meets will move his front door every day unless he pays > tripple rent. > > I suppose the question can be phrased more theoretically as: > > "If a LIR is obliged to assign address space (is it?), No, it is not. An IPS's LIR may have very good reasons (financial, security, AUP...) *not* to assign addresses to an entity wanting to become connected and obtain services, and even become a customer :-) > wouldn't it make > sense to oblige a provider to route it" No. That's what contracts are invented for, either implicitely by subscribing to a service or explicitely by negotiating ToS/QoS. > or the other way'round: > > "If an address space request is made, is the non-willingness of an ISP to > route it sufficient grounds to deny the request?" I think so. Unless there are other valid reasons for an LIR to perform an address assignment for a customer or peer (i.e. PI space, that is *not* accepted by the IPSs routing layer, but may be used on other links of the customer). > > Well, there is no formal policy obligation for you to route the addresses > > you assigned, but what would those addresses then be good for? > > That's the point. Can the customer somehow (e.g. by submitting a ripe-141) > force the ISP to assign him a static ip address? Or will it get him nowhere? It depends what the definition of "customer" looks like in that case. When the ISP does offer a package that includes the privilege to obtain service from the ISP's LIR, then the customer can probably (try to) use force (legal measures or refusing to pay fees). Trying to throw a 141 form at some free dial-in service provider would probably "get you nowhere" :-) But in general I would try to find a different ISP... ______________________________________________________________________ =Hi, = = This is a very interesting problem. = = As far as I am aware there is no obligation to assign IP space =when someone becomes a customer of an entity served by a LIR. Again, I share this view. = Possible problems when submitting a RIPE-141 document: = =- There is no obligation to process the request True =- The LIR may charge for processing the request True =- The LIR may charge for the assignment Not really. This is seen as "selling" addresses! =- The provider may charge for the routing True = You could opt to ask RIPE for a PI allocation, You could find an LIR that is willing to process PI requests, either for a charge or for free, or you can build your own LIR ;-) =but noone is required to route it for you. True, and for both, PI *and* PA =Regards, = =- marcel Regards, Wilfried.

2 1

IPv6 representation in perl
by Guy Vegoda 26 Feb '01

26 Feb '01

On an entirely different point, I would like to start a separate thread about the best way to represent IPv6 numbers in Perl. I have the following thoughts. Possible ways may be: (1) Using a 128 element binary array, and writing all the binary maths yourself to manipulate that. (2) Using a 4 element array storing four hex characters from 0 to F. Probably far slower than the above method, but smaller for storing many IPs. I know that Perl can natively think in hex, so maybe it would be faster than I think. (3) Having a complex structure, where the IP number is stored in parts - the top /48 (in any of several formats), with variable structure subnets underneath that. Could be tricky for the maths, if the variable bits are badly thought out. (4) Storing four "IPv4-esque" decimal strings in an array, or as part of an object. Then do the maths on each of those separately. Probably fast. (5) Storing a single parant /48 object (in any format) and have child objects with only the bottom parts of the IP to worry about. I wonder how that would work. -------- When I have gone away and read more of Mastering Algorithms with Perl, I will come back and maybe make some more comments. Just as a disclaimer - I am in no way making any comments on Manuel's already excellent libraries. I have already spoken to him about them, and would not want to appear to "go behind his back". This is an entirely intellectual exercise only. In fact, I would very much like to see Manuel post his own thoughts and real world experiences on the subject (of which I have none). I wish every one well, Guy -- Guy Vegoda \ guy(a)vegoda.org *Please do not send html* NIC: GUY-RIPE \ guy(a)cryptography.org.uk *attachments* Unix, Linux Hobbyist \ +44(0)20 7961 8318 (work) www.thenakedfrenchman.com \ +44(0)958 469 532 (cell)

2 1

Re: IP Management Tool - Minimum Requirements
by 'guy＠sirius-cybernetics-corporation.com' 26 Feb '01

26 Feb '01

> Yes I know about asused. It is a good tool for determining overlap > status or some missing attributes/contact informations and to get some > statistical information. I think it will be usefull to access asused from > IPMT. But my main concern was also whether we can find such a > criteria/metric that can be used both RIR's and LIR's ? Here are some > metrics I have thought; > > - Possible largest allocation left in the totall allocation of LIR > - How fragmented the allocation space (may be mapped to above one) > - How summarizable are the address space assigned to a customer. > - How summarizable are total allocations ? > > > Summarizability issues seem to conflict with RIR policies. However > I think that it is a fact of routing world. > > Regards > arif I do not think there is anything wrong with detailed reporting. I think that maybe the program needs a specification before we can really use metrics to measure anything. I too would like to see IPMT eventually give verbose output on the state of the registry. However, will this not to some extent duplicate functionality of programs such as asused-public that already exist? Please clarify what you would like, because I think I am still confused. I think that Maldwyn's comments would be useful here. Guy -- Guy Vegoda \ guy(a)vegoda.org *Please do not send html* NIC: GUY-RIPE \ guy(a)cryptography.org.uk *attachments* Unix, Linux Hobbyist \ +44(0)20 7961 8318 (work) www.thenakedfrenchman.com \ +44(0)958 469 532 (cell)

1 0