Have you requested a block and had trouble?
Scott, Well, ummm, yeah. We recently asked for a 256 address block of Class C space and was told to go to ANS for addresses. We've currently got a 256 block that's about used up, plus we obtained another 256 block for a customer of ours who had burned their B. We were asked to prove that we had used up both blocks (and proving that we had) were then told to go to ANS. I can put in arguments here about supernetting, routing, and multihomed networks. It doesn't make sense to me for a network that is multihomed to have a portion of someone else's address space... We'd thought about just asking for a 1024 address block, but given our recent experiences (we really got grilled on the second 256 block ["are you using those numbers"]), we decided not to even try. henry
Henry,
Have you requested a block and had trouble?
Scott,
Well, ummm, yeah. We recently asked for a 256 address block of Class C space and was told to go to ANS for addresses. We've currently got a 256 block that's about used up, plus we obtained another 256 block for a customer of ours who had burned their B. We were asked to prove that we had used up both blocks (and proving that we had) were then told to go to ANS.
Service providers are coming out of the wood works. The size varies drastically bringing up the question of what a service provider is with respect to CIDR. Jon Postel and I have been discussing this issue and will try to better define the guidelines. I am sorry that you got caught-up in that effort. I think that Jim S. has contacted you to straighten this out. He was unaware that you were multihomed based on your response to his questions.
I can put in arguments here about supernetting, routing, and multihomed networks. It doesn't make sense to me for a network that is multihomed to have a portion of someone else's address space...
I agree. We have been asked to coordinate large allocation of blocks with the next level service providers. In your case ANS and ???.
We'd thought about just asking for a 1024 address block, but given our recent experiences (we really got grilled on the second 256 block ["are you using those numbers"]), we decided not to even try.
Do you expect to grow enought to justify the allocation of 1024. If so please send in the request and Jim will discuss it with you.
henry
Scott
From my reading of the Rekhter & Li CIDR paper I believe the allocation strategy is to assign blocks to immediate providers ("retail" or "access"?) and that those providers get their allocation from the regional NIC, which for this purpose would be the InterNic acting as
Scott, I don't think network service providers should take address blocks hierarchically related to allocations from other service providers in the US. I don't believe this is an issue of multi homing, it is simply the issue of minimizing the probability of customers having to change their addresses. If an access provider changes its "long haul" provider it would be nice if there were no pressure for the customers of the "access" provider to change addresses. the NIC for the U.S.
From a U.S. point of view I could see carving the U.S. up into a few regions (probably a max of 4) where immediate/access/retail providers for a restricted region (provider XXX is a dialup provider for the NY area) get their allocation out of those geographically defined blocks rather than a generic undifferentiated US block.
cheers, peter
Peter, First someone must provide the definition for : What is a provider? Do we just give a block to any one who said they are a provider? This question has been batted around over and over with no answer. It sound simple but we get some very small operation asking for their own block. Shouldn't their current next level provider be involved to some extent with these smaller operations? Scott
Scott,
I don't think network service providers should take address blocks hierarchically related to allocations from other service providers in the US. I don't believe this is an issue of multi homing, it is simply the issue of minimizing the probability of customers having to change their addresses. If an access provider changes its "long haul" provider it would be nice if there were no pressure for the customers of the "access" provider to change addresses.
From my reading of the Rekhter & Li CIDR paper I believe the allocation strategy is to assign blocks to immediate providers ("retail" or "access"?) and that those providers get their allocation from the regional NIC, which for this purpose would be the InterNic acting as the NIC for the U.S.
From a U.S. point of view I could see carving the U.S. up into a few regions (probably a max of 4) where immediate/access/retail providers for a restricted region (provider XXX is a dialup provider for the NY area) get their allocation out of those geographically defined blocks rather than a generic undifferentiated US block.
cheers,
peter
Claudio & I have been hashing that one around since the Ohio meeting. I have a rough draft somewhere around here that should help define what might be considered an NSP. -- Regards, Bill Manning bmanning@rice.edu PO Box 1892 713-285-5415 713-527-6099 Houston, Texas R.U. (o-kome) 77251-1892
Attached is a brief review of what Claudio & I cobbled together. I don't remember if all the comments were added, and conversations today point out the need to either restrict the scope to IP (the wording of the current draft seems to point to an IP-centric view), or widen the view to support multi-protocol view(s)....:-) Additional, -CONSTRUCTIVE- comments are welcome. What is a Internet Provider? An entity that provides IP service to a customer base, either in a closed environment (no external world view) or open, world connected Internet service. At a minimum this would entail a bilateral aggreement with another provider to exchange routing information between the two. It would also imply that the instant any bilateral agreement is made, that one of the common points of this agreement is to recognise the IANA for what it is, and agree that the documents produced by ISOC standards and published in the offical RFC repositories are the standards by which compliance with an IP network are defined and form the sole basis for arbitration of disputed points. RFC change occurs in the normal fashion as identified by the ISOC. This does not imply any actual application support between providers. Areas that may define base level customer service are DNS registration, and some SMTP configuration support. Additional user services for clients are infotools like archie/ftp/gopher/www. Higher level services like help desk and training may be added. It may be appropriate to require some level of periodic accounting/checkup for all providers by the IANA for numbers allocated to providers. It may also be required that a provider must be refered by an existing provider. This seems to cover the base set of what an Internet provider does and may do. Additional requirements may be imposed by other entities that have a controlling interest in one or more providers. It seems that a checklist of items that when reviewed, are generally positive, indicate that internet provisioning is occuring, is a useful tool. -- Regards, Bill Manning bmanning@rice.edu PO Box 1892 713-285-5415 713-527-6099 Houston, Texas R.U. (o-kome) 77251-1892
Attached is a brief review of what Claudio & I cobbled together. I don't remember if all the comments were added, and conversations today point out the need to either restrict the scope to IP (the wording of the current draft seems to point to an IP-centric view), or widen the view to support multi-protocol view(s)....:-)
Additional, -CONSTRUCTIVE- comments are welcome.
What is a Internet Provider?
An entity that provides IP service to a customer base, either in a closed environment (no external world view) or open, world connected Internet service. At a minimum this would entail a bilateral aggreement with another provider to exchange routing information between the two. It would also imply that the
Seems that the above is a sufficient definition.
instant any bilateral agreement is made, that one of the common points of this agreement is to recognise the IANA for what it is, and agree that the documents produced by ISOC standards and published in the offical RFC repositories are the standards by which compliance with an IP network are defined and form the sole basis for arbitration of disputed points. RFC change occurs in the normal fashion as identified by the ISOC.
I do not believe that either IANA recognition, nor anything with RFCs, and certainly nothing whatsoever related to the ISoc has *anything* to do with whether someone is a service provider or not.
This does not imply any actual application support between providers. Areas that may define base level customer service are DNS registration, and some SMTP configuration support. Additional user services for clients are infotools like archie/ftp/gopher/www. Higher level services like help desk and training may be added.
Unimportant to the definition.
It may be appropriate to require some level of periodic accounting/checkup for all providers by the IANA for numbers allocated to providers. It may also be required that a provider must be refered by an existing provider.
That's fine, but is not related to a short and crisp service provider definition.
This seems to cover the base set of what an Internet provider does and may do. Additional requirements may be imposed by other entities that have a controlling interest in one or more providers. It seems that a checklist of items that when reviewed, are generally positive, indicate that internet provisioning is occuring, is a useful tool.
Also not sure why this is necessary and why the definition needs any wording like imposed, controlling, and such are needed. May be one way to deal with that is to have a clear/crisp/short definition. Separate from that could be a suggested document about suggested guidelines for interactions swith other SPs which could include some of the rest of Bill's text. Hans-Werner
An entity that provides IP service to a customer base, either in a closed environment (no external world view) or open, world connected Internet service. At a minimum this would entail a bilateral aggreement with another provider to exchange routing information between the two. Seems that the above is a sufficient definition. It's either too much or not enough: - "An entity that provides IP service to a customer base" in itself sufficiently defines an internet (lowercase!) service provider. - Even though a particular customer of a certain internet service provider may have access to most of the Internet (uppercase!), certain parts of the Internet may well be blocked to him due to AUP constraints. - Exchanging of routing information is pretty useless if it isn't accompanied by exchanging of traffic. In fact it's *only* the latter that determines whether a certain internet service provider is an Internet service provider. - One of the parties in the last sentence must have Internet connectivity, otherwise we're still speaking of internet service providers only. So I would propose the following wording: An entity that provides IP-based services to a customer base, in a closed group of networks or - as far as possible within Acceptable Use Policy constraints - to the Internet at large. At a minimum the latter would entail a bilateral agreement with at least one Internet-connected service provider to exchange traffic and routing information between the two. Piet
From: William Manning <bmanning@is.rice.edu>
What is a Internet Provider?
An entity that provides IP service to a customer base, either in a closed environment (no external world view) or open, world connected Internet service. At a minimum this would entail a bilateral aggreement with another provider to exchange routing information between the two.
Bill; I think "coordinating" is the operant adjective. Anyone who needs to coordinate is de facto a coordinating internet service provider. Sounds circular, so focus on the adjectival description: What is a Coordinating Internet Provider? ------------ A coordinating internet service provider is a ISP that needs to coordinate with other ISPs for information exchange and service continuity to support the global Internet within the context of <appropriate> standards and <appropriate> practices. That pretty much lets anyone in who agrees to participate. --Kent
scottw@internic.net (Scott Williamson) writes: Peter,
First someone must provide the definition for : What is a provider? Do we just give a block to any one who said they are a provider? This question has been batted around over and over with no answer. It sound simple but we get some very small operation asking for their own block. Shouldn't their current next level provider be involved to some extent with these smaller operations?
Scott
FYI here is what we do at the RIPE NCC: - We ask them for their own estimates on what they will need - immediately - within one year - within 2 years We ask for estimates whith some justification. This usually gives us some idea of what they are about. - If the above does not give us enough information we will use any sources available to us. - If we conclude that delegating a whole block is not justified at that point we will delegate part of a block. After all a block is not a physical quantum. - Crucial: We always tell them more than once that this is just the initial procedure and once they come back with a report on how the first allocation was used they can readily get more address space. Most importantly: We will, as a matter of policy, not get involved in any discussion whatsoever about who is a provider and who is not. This will just lead to fruitless discussions and cycles wasted. Of course all providers are expected to contribute to the funding of the RIPE NCC :-). Hope this helps Daniel
participants (8)
-
bmanning＠is.rice.edu -
Daniel Karrenberg -
henryc＠oar.net -
hwb＠upeksa.sdsc.edu -
Kent England -
peter＠goshawk.lanl.gov -
Piet.Beertema＠EU.net -
scottw＠internic.net