Hi Michael,

...

Umm, why not? And how do you _know_ that it is a "few"?

Because the RIRs don't have the mandate or the resources to cater to a few companies who have gotten themselves into a fix. I know that it will only be a "few" because the IPv4 supply is shrinking dramatically down to zero. Soon there will be very few IPv4 blocks available to be transfered therefore the companies (sellers and buyers) who would benefit from a transfer policy are very few.

This transfer policy is not about making sure everybody has all the addresses they need, but to make sure that the addresses that are allocated but unused can be used by other organisations. We don't know exactly how many addresses are allocated but unused, so how can you determine the number of organisations that will benefit from this policy? It is also not true that we can't make policies for a small group of users. We have for example a policy for Internet Exchange Points. I think we should try to avoid special-case policies, but this proposed policy is not special-case. If organisations choose not to transfer their resources, that is ofcourse up to them.

Also, how do we know that sellers and buyers would actually benefit, in a real sense, from a transfer policy? All the buyers will be companies in a desperate situation, and as soon as the transfer becomes known in the RIPE database, their customers could begin to flee.

I think that you over-estimate the number of people that look at the RIPE database when choosing their ISP. And if organisations think it will harm their image, they can always choose not to transfer any address space.

...

I'm afraid this is a completely bogus prediction about the action of courts, made by a non-lawyer, with no specification of which legal system/courts he is talking about, nor the legal grounds upon which such a decision would be based, or what kind of litigation would put the issue into the courts to begin with.

This applies equally to your comments and to those of almost everybody who is involved in RIRs and transfer policy discussions. This is a good reason for us to reject these transfer policies so that RIPE can focus on its competencies, not try to take on the legislative and economic functions of government departments.

The management of address space _is_ the primary competency of RIPE... The RIPE NCC also checks all policy proposals to determine if they cause any problems, so expect their lawyers to look at what we decide and give feedback where neccesary.

...

The auctioning of mobile spectrum in the U.S. has not forced broadcasters, who got their spectrum for free, to account for their spectrum holdings according to a projected market value.

Thanks for making my point clear. Governments, and their regulatory agencies, have the power to do things which RIPE cannot do.

This has nothing to do with power but about the scarceness of a resource. Scarce resources become valuable, other resources don't. The existence of a transfer policy won't change the value. If the value is high enough, companies will find a way to trade it, even if it is not officially allowed...

...

This is just weird. If v4 addresses are sold in a competitive market, you call them "unpredictable." But you call the price of v6 service "predictable" because....it is sold in a competitive market.

A small market with few buyers and sellers, is very unpredictable. You need a certain amount of trading volume for the law of large numbers to kick in and begin to get some level of predicability.

If the price of a resource is too unpredictable or too high for an organisation, they can always choose not to pay it and search for another solution. This transfer policy gives them that choice.

...

Sure, they'll sell access to 1/10th of the internet for less than the price of 100% access. What you may not be taking into account is the universal connectivity associated with v4.

I'm afraid you don't understand the technology that underlies all of this. There is no good reason why IPv6 network access could not provide reachability to 100% of the Internet. The IPv6 transition has been going on for 10 years or so and most of the technical issues have been worked out. Things like 6to4 and Teredo are well understood and widely used today. NAT-PT is also in use, and the IETF is working on resolving the last design issues with their work on NAT64.

For the client-side this might work, but what about organisations that offer services? If those services won't be reachable over IPv4 they will have a serious problem.

I am assuming that ISPs will sort out the remaining technical issues and deploy them within the next two to three years so that they can indeed provide IPv6 network access that allows communication with 100% of the Internet.

I hope so... Sander

a) stop growing, turn away customers (or, more likely, raise prices to drive out the lower paying customers to free up space for new customers)

b) buy up any organization they can find that has address space

c) go to the black market to obtain new address space at a market- defined price

d) accept new customers only if they come with their own address space

e) start selling (potentially multi-layer) NAT'd connectivity

Do you see any other options?

Most ISPs are large enough that they sell several different products to different customer bases. I expect that it will be more common to cannibalize less profitable product lines to get the IPv4 addresses needed for more profitable products. It's not the product price that is the factor because lower paying customers may be more profitable in aggregate. For instance if you have two broadband products, one which assigns a single IPv4 address using DHCP and one which provides a static /29, then the static product which has the higher price, may be the one which is cannibalized because it is not as profitable as leased-line sales using a /29. If you call this option f) then I would rank them in popularity: f) d) e) b) c) a) Choices f), d) and e) have a lot in common, i.e. they apply ingenuity to work around the problem rather than banging their heads against the brick wall of IP address shortage. Also, if you are enumerating choices in this way, it is devious to not include variations of deploying IPv6. Also, I expect that ISPs will come up with some variation of moving the NAT function into the ISP where they assign unregistered addresses to their customers, and then NAT this at a gateway on the ISP premises. With this variation, ISPs will reuse addresses that are registered to geographically distant organizations with whom their customers are unlikely to need to communicate. Since this provides a very large pool of IPv4 addresses that can be used more-or-less as RFC 1918-style addresses, there is enough flexibility to avoid problems. For instance, you could offer a range to a customer telling them who is registered to use that range, and if they veto it because they might need to communicate with that region, then you can pick another range. Perhaps we should not be talking about the brick wall of IPv4 exhaustion because it's more like a hall of mirrors. Some people will take their chances on navigating the hall and finding an exit, others will prefer to avoid the fun and just deploy IPv6. --Michael Dillon