Hi Daniel, I hope I can clarify some of your points below: At 03:37 AM 8/21/2003 +0200, Daniel Roesen wrote:

On Wed, Aug 20, 2003 at 04:24:19PM +0200, Dominic Spratley wrote:

...

The RIPE NCC is pleased to announce the release of new request forms for IPv4, IPv6 and AS requests.

Uhm, can someone point me to the discussion and decision finding process leading to these rather radical changes to especially the PI request? I must have missed those.

We hope that the new forms will improve services to our members by only asking for relevant information relating to the evaluation of a request. By trying to be explicit about what information Hostmasters need, we aim to reduce the number of E-mails asking for further information and therefore reduce the overall time it takes for our members to receive their resources. We considered this to be an action on us from RIPE 38: "< 38.2 NCC Implement changes to the form while taking into account the comments from the WG >" http://www.ripe.net/ripe/wg/lir/lir-actions.html

In detail, I'm somewhat _shocked_ about things like:

- "why-pi:"

This is a new requirement. Formerly, you didn't have to justify that.

You raise a good point which we need to clarify. We are not asking members to justify to us why they need PI address space. However, what we find is that a number of LIR's misunderstand the difference between PA and PI address space and that time is often lost in trying to establish if this is the actual type of resource the LIR wants. By asking them to explain why they need PI address space rather than PA address space it helps resolve this issue and the request is processed faster. Having read through the supporting notes for PI assignment request form, I agree with you that it is ambiguous to say the least. We will update the document in the next release.

In one example, "Also security reasons means independence is required." is used to justify PI. Can anyone explain what "PA or PI" has anything to do with security?

In our examples we tried to use the terms and phrases that our members actually use in the forms that they send in to us. Statements like the one above are regularly seen, in conjunction with further supporting information, in PI request forms.

- "Respecting the goal of address space conservation, it is not acceptable to request a larger number of IPs than are needed for routing or network administration."

Any LIR is allowed to do so by granting them an allocation. For LIRs, address space conservation is less of an issue? _Especially_ enterprise LANs/WANs can have some very complicated routing situations requiring sensible subnetting.

RIPE - 127 talks about: "Assignment criteria for both kinds of address space will be exactly identical with regards to the amount of address space assigned, the registration requirements, etc. This also implies that assigning PI space prefixes longer than 24 bits is perfectly acceptable if the request does not merit 8 bits of address space to be assigned." http://www.ripe.net/ripe/docs/pi-pa.html and in fact it is this particular part of the policy that sparked off the PI-TF and is currently under discussion on the address-policy-wg mailing list.

- [EQUIPMENT DESCRIPTION]

This requirement _can't_ be meant serious. We're not on April 1st, are we? The silliness of such information requirement was already recognized within the IPv6 allocation realm, so why introducing it now again for IPv4?

This is a difficult one as it does require the LIR to do more preparation on their side. The less time a Hostmaster spends sending E-mails asking for additional information, the sooner the LIR gets their requested resource. Asking for the type of equipment used really helps the Hostmaster understand what and why IP addresses are being requested.

Further, the term "peer" is used in a very unclear way in the ASN request form and the supporting notes.

Formerly, one uplink and one peer (commercial term) were enough to justify an ASN, as this constitutes a "unique routing policy". The ASN request form allows this interpretation of "peering partners" with "peering" in the technical sense. BUT the supporting notes specifically mention the requirement for two peers to "ensure that the organisation is multihomed". For me, multihoming means to have at least two _upstreams_. This would be a more stricter policy than before. What was the real intention behind these phrases? These documents should be very specific in the usage of "peer". It might be interpreted technically or commercially, each implying different constraints. I suggest changing the wording from "peer" to "upstream" in order to avoid ambiguities.

Again this is a good point. However I don't think that either term is appropriate for all situations. When writing these documents we used the word peer as we wanted to emphasise it was the routing policy which was important. When designing these new forms we made sure that they could be updated and amended given feedback from the community at a later date. We appreciate your comments and hope that we have been able to clarify at least some of your points and address others in future releases. Best regards Dominic Spratley Registration Services Manager RIPE NCC

Hi, I agree with Daniel and Sascha, these issues are very important and needs to be discussed! As Daniel points out, these radical changes haven't been discussed in the maillinglists before beeing implemented - why? It seems reasonable enough to update the old forms, but I cannot see any big advantages in the new versions? Presently it is rather complex and time consuming to gather all the information needed to apply for IP addresses, especially since most customers do have other things than Internet on their mind, so they have to be given a very thorough explanation in order to make sense of such an application. The question is what information is needed and why? As a general rule, endusers are allowed to expect a maximum increase of 100% pr. year of IP addresses - so, to find out the number of IP addresses to assign the enduser simply needs to tell how many nodes requiring IP addresses they have, in case the endusers IP address requirement is higher than the number of nodes, a detailed technical documentation justifying this need is required. I don't see much reason for going into details about the organizational structure of the enduser neither the reason for requesting detailed information about each node.. either its there or its not! Of course there are situations when enduser informations sounds rather unrealistic, but if an enduser intentionally lies about their need in order to get more IP addresses than justifable, should the LIR be responsible for this? Regarding PI-space, in case an enduser can justify for example a /24, unless the enduser intends to be become LIR, should there be any reason for the enduser not to get a PI assignment if this is the endusers request? About the new forms, I assume the "ADDRESS SPACE USER" is intended to replace Organization overview so it is no longer required to explain the organization of a enduser in detail? If so, I think this is a good idea. "Does the organisation already have address space that ...." The only reason I can see is if the enduser has a connection to the Internet and buys another, probably on another location (of course the enduser could request extra IP addresses unjustifably, if so, the LIR should not waste time filling out the form but just explain the situation to the enduser), could there be other reasons for this? If not, I think the question should be clarified if not omitted. The whole address-space-returned-situation as well as stating exactly which IP nets the enduser has are a bit tricky.. Sometimes the enduser does not know exactly what IP addresses they have (or forgets to mention it..), the LIR has no way to be sure that the received information is correct (especially if the LIR, at which the enduser is also customer, has not registered the IP net). Who's responsibility is it that IP addresses are returned? I would think that it might be the LIR registering the IP assigment, but I haven't been able to find it? If so, why? #[EQUIPMENT DESCRIPTION]# % % Please describe the equipment that will be used in the % network. Indicate the function of the equipment and provide % information regarding the way it uses IP address space. equipment-name: manufacturer-name: model-number: other-data: What is the reason for requesting the model-number and the other details!??! According to Dominic's responds it seems that this information is necessary for the hostmaster to evaluate the request, but how does it help to find out the number of IP addresses to assign if you know if a node is manufactored by IBM/HP/whatever? Or if is a model with 1,4 GHz CPU or 1,7 GHz? I don't believe there is any reason for this, if others could agree the hostmasters could save a lot of time by not having to look at these details. As both Daniel and Sascha has said it is simply not reasonable to request a network diagram, of course it can be optional. Dominic says that a reason for the thorough questions regarding PI-space is that some LIRs do not understand the difference between PA and PI!! But how can this justify consuming time at all the other LIRs?! But I don't understand how you can be LIR without understanding PA/PI? You have to state that you have read the relevant documents in order to become contact person, if PA/PI is not clear after reading the documents then the documents of course needs to be updated - but it doesn't make sense to have LIRs assign IP addresses without understanding the basic principals. I agree with Sascha that its not acceptable to have requests delayed because the forms requires these unreasonable informations! If other LIRs do not see this as a problem simply because of their large AW it doesn't make much sense; even if you assign most of your requests without consulting a Ripe hostmaster you're still required to have the information for the relevant form available... Anyway, I do not understand who benefits from having irrelevant information gathered! I hope somebody have comments to the above and can explain the reason for these to-be time consuming tasks, if not, these forms should be radically simplified. Med venlig hilsen/Best regards Christian Rasmussen Hosting manager, jay.net a/s Smedeland 32, 2600 Glostrup, Denmark Email: noc@jay.net Personal email: chr@corp.jay.net Tlf./Phone: +45 3336 6300, Fax: +45 3336 6301 Produkter / Products: http://hosting.jay.net

-----Original Message----- From: address-policy-wg-admin@ripe.net [mailto:address-policy-wg-admin@ripe.net]On Behalf Of Sascha Lenz Sent: 25. august 2003 11:43 To: Dominic Spratley Cc: ncc-services-wg@ripe.net; address-policy-wg@ripe.net Subject: [address-policy-wg] Re: [ncc-services-wg] Request Forms: updated and available on LIR Portal

Hay,

On Thu, Aug 21, 2003 at 04:02:16PM +0200, Dominic Spratley wrote:

...

Hi Daniel,

I hope I can clarify some of your points below:

actually i didn't really want to get into this discussion, but it looks like _noone_ else has anything to say about the issues Daniel raised? Than can't be since he is right in most points.

...

At 03:37 AM 8/21/2003 +0200, Daniel Roesen wrote: [...]

most points have been more-or-less clearified by Dominic's answer, but:

...

...

- [EQUIPMENT DESCRIPTION]

This requirement _can't_ be meant serious. We're not on April 1st, are we? The silliness of such information requirement was already recognized within the IPv6 allocation realm, so why introducing it now again for IPv4?

This is a difficult one as it does require the LIR to do more preparation on their side. The less time a Hostmaster spends sending E-mails asking for additional information, the sooner the LIR gets their requested resource. Asking for the type of equipment used really helps the Hostmaster understand what and why IP addresses are being requested.

... i still have a problem here if that should be a "must provide this information".

I have no much problems with questions like "why-pi" or "routing-reasons" because they can obviously be answered with standard phrases (which renders them useless again, but if RIPE NCC sees that many "bad" requests from many LIRs (aren't they trained?), it's OK for me, no much work, no stupid questions).

BUT... information like the equipment being used or a network plan or so _must_ be optional information, not _required_ information. It can't be that a company is _required_ to make statements about their network and equipment if they want to have IP-Adresses! (and no, RIPE doesn't always qualify for receiving such information).

Besides that it's way too much work to fill that in seriously for big requests, and doesn't really make sense for small requests.

This should be _optional_ information in cases where it's needed to explain an unusual request, whatever, "we have only one machine but it has 500 IP Adresses due to this and that special purpose which can't be handled otherwise due to these and those hardware and software limitations".

Is it really only me who is of that opinion, or noone recognized this problem or doesn't think it's a problem? Or isn't it meant as _required_ information anyways and a "N/A" is enough in most cases as answer to this part of the request? ...curious.

The reason i noticed that this might get a problem was due to the unholy

[...] #[REQUIRED INFORMATION]# [...] % % 2. Please provide a network diagram in PostScript or JPEG % format showing your IPv6 network. Submission instructions % can be found in the RIPE document "Supporting Notes for the % Initial IPv6 Allocation Request Form in the RIPE NCC Service % Region" found at: % % http://www.ripe.net/ripe/docs/ipv6-support-initial.html % % The diagram should indicate expected completion dates of % construction as well as how much IPv6 address space is % expected to be used in each subnet. %

% [...]

in the (old and new) IPv6 Initial Allocation request form.

Since it says "required information" there, i already assumed that it might get a problem to get the IPv6 Allocation for us without providing this one. But i didn't see the point in doing so if i can make our network plans clear in text form, and explain the reasons why a network diagram for an IPv6 network is _completely_useless_ in general at the moment and especially in our case.

Though, even on my 2nd mail on that request i just got the reply from some RIPE Hostmistress that "the inability to provide this network topology map calls into question the need for the IPv6 addresses being requested.".

At the moment i'm really mad about this. We are currently prevented from deploying IPv6 addresses immediately because i don't have any Windows PC with Viso or something at hand, let alone the time to paint a very useless topology of our IPv6 overlay network, though i explained it in detail in text. (Not to mention that the statement that an active LIR which assigns IPv4 addresses for years is questioned that it will assign IPv6 addresses just because it doesn't like to waste time on painting network maps of some routers and tunnels is VERY doubiously to me.)

Why the hell is such information _required_? Information about what equipment being used and a network topology map on a picture might be additional information at best.

I fear that IP(v4) request will be denied or at least delayed in future if the RIPE Hostmasters are gonna complain about missing or unsatisfying information about the equipment in the future, too.

That's not acceptable i think. But i might be wrong, we're realatively small and don't have that many requests in general, i just wonder why noone of the "big LIRs" sees this problem? You all have AWs that big that you don't care about filling in RIPE forms anyways and have payed managers to paint network diagrams all day? *hint* :)

...

...

Further, the term "peer" is used in a very unclear way in the ASN request form and the supporting notes.

Formerly, one uplink and one peer (commercial term) were enough to justify an ASN, as this constitutes a "unique routing policy". The ASN request form allows this interpretation of "peering partners" with "peering" in the technical sense. BUT the supporting notes specifically mention the requirement for two peers to "ensure that the organisation is multihomed". For me, multihoming means to have at least two _upstreams_. This would be a more stricter policy than before. What was the real intention behind these phrases? These documents should be very specific in the usage of "peer". It might be interpreted technically or commercially, each implying different constraints. I suggest changing the wording from "peer" to "upstream" in order to avoid ambiguities.

Again this is a good point. However I don't think that either term is appropriate for all situations. When writing these documents we used the word peer as we wanted to emphasise it was the routing policy which was important.

So, the policy didn't change to "you have to have two _upstreams_ to get an ASN" but it's only a language-confusion here?

P.S.: I think this discussion belongs to the address-policy-wg, not ncc-services-wg? Sorry for crossposting, remove the latter in any replies if i'm right.

-- ========================================================================== = Sascha 'master' Lenz SLZ-RIPE slz@baycix.de = = NOC BayCIX GmbH = = http://www.noc.baycix.de/ * PGP public Key on demand * = ================================================================== ========