Factors, actions influencing the possibility/timing of IDR for IPv6-basedrouting domains?

Ted Mittelstaedt tedm at ipinc.net
Mon May 18 20:23:18 CEST 2009


> -----Original Message-----
> From: Tom Vest [mailto:tvest at pch.net] 
> Sent: Friday, May 15, 2009 7:12 PM
> To: Ted Mittelstaedt
> Cc: 'IPv6 Ops list'
> Subject: Re: Factors,actions influencing the 
> possibility/timing of IDR for IPv6-basedrouting domains?
> Hi Ted,
> Thanks for the reactions.
> I'm going to attempt to rephrase slightly to fit them into 
> the framework that I attempted to describe in my original request.
> Please let me know if I've missed or misinterpreted anything.
> Some factors/actions that you cite:
> 1. RIRs will attempt to recover some IPv4 for redistribution.
> 2. Some operators will undertake internal "rationalization" 
> of their own IPv4 assets.
> These are interesting, but not directly relevant to my 
> question about prospects for native v6 IDR (except perhaps as 
> potential confounding/ delaying factors).

You had asked what actions operators of other routing domains
could take that would affect standalone IPv6 routing becoming
possible.  If operators of other domains refuse engage in IPv4
recovery, IPv6 will come a lot faster.

I think if you look at the math of it, this is much more of a
major issue than most people think.  Right now only about 1/2
the IPv4 allocated is advertised in the dfz, if aggressive
transfer sales and reclamation take place, in conjunction with
heavy use of NAT, it could potentially stall the IPv6 effort

Also if end users accepted private IPv4 from the ISP as a matter of
course, I think it would be a very big disincentive to rolling
out IPv6.  And end users are
not known for being very smart about networking.

It's too early to tell what impact that IPv4 reclamation will have,
some people say a lot, some say a little.  I do think that
this is an area where perception becomes the reality - I think if
the perception is that IPv6 is adopting very slowly, that
reclamation efforts will be pursued with much more vigor and
will yield a lot more fruit.  By contrast if perception is IPv6
uptake is fast, then reclamation will not be vigorously pursued
and will die away.

Another thing too about the transfer market is that if perception
is IPv6 is adopting slowly, then even though reclamation will
be vigorous, orgs doing the reclaiming will keep the IPv4 to
themselves.  This will of course prolong IPv4 usage and make it
a lot harder for new entrants to get IPv4 numbering.

I also think that the amount of freed-up IPv4 an ISP can
create (for resale or use with new customers) is proportional to
the amount of effort they spend on reclamation.

> Others that you mention:
> 3. Some IPv4-based operators will sell IPv4, and some 
> aspiring new entrants will buy IPv4.
> 4. Mobile access providers will increasingly use IPv6 to add 
> new customers and services.
> (3) is a perfect match with my original list, and (4) is a 
> special case of another factor that I cited initially.
> You also suggest that:
> 5. Content providers will increasingly "connect to the IPv6 network."
> This one is ambiguous. Are you suggesting that online content 
> providers will build out their own wide area native-IPv6 
> distribution platforms to reach native IPv6 wireless/mobile 
> access customers? Are you anticipating other sources of 
> demand for native IPv6-based content
> -- i.e., other than recently added wireless access customers? 
> Are you anticipating other distribution mechanisms (e.g., 
> other than DIY) for getting native IPv6-based content to 
> those new native IPv6 sources of demand? Presumably not all 
> online content providers will have the scale or means to 
> build their own national/international infrastructure platforms...

No.  What I am saying is that the large content providers - Google and
the like - are already onboard with this IPv6 deployment and are
working on becoming dual-stacked now.  It is not to capture more
customers as much as it's a community service effort, at this point.
To those providers, getting IPv6 compliant is a money-saver in the
long run because they know that ultimately they are going to have to
spend the development dollars to get their stuff compliant, and
if they get complaint now, then in the future when they have built out
more stuff they will have less stuff to change, and it will be cheaper.

What I mean more are the smaller content providers.  For example, go
to Google, type in "barbeque" and you will get about 100 links to a bunch
of small time people who are selling bbq sauces and suchlike, and
providing content such as recipes and cooking tips to attract people
to their sites.

Just about all of these folks are hosting on one of the large 
virtual server farms, run by one of the virtual server companies
on the Internet.  None of those people are going to go to their
server farm operators and demand the operator add IPv6.

But, those farms know the same thing Google knows, which is that
eventually there's going to be IPv6-only clients out there, and
they don't want those clients complaining to their own customers,
and their own customers perhaps leaving without even saying anything.
Plus, you never know exactly what Google is going to use as search
criteria.  Google is spending big bucks on their own IPv6 deployment,
and it wouldn't be farfetched to imagine that Google might one day
quietly slip "Reachable by both IPv4 and IPv6" on to their list of
items that increase your page ranking.

You have to take it as fact that this is going to be a transition - 
first IPv4-only, next dual-stacked, last IPv6-only.  And likely it
will be on a classic bell curve.  In the beginning the number of
dual-stacked sites will be small, but when it hits 20% then everyone
and their dog will climb on, then the rate of adoption of dual stack
will peter off after it hits 80%.  Then the rate of IPv6-only will
follow the same curve.

The content providers will be further along this curve because nobody
wants to be NOT dual-stacked when the first customer with money in
their pocket that is IPv6-only hits the Internet.

Then the content providers will be the LAST on the curve to shut down
dual-stacking and go to IPv6-only, for the same reason.

> And finally:
> 6. Somebody (IPv4-based operators? IPv6-based new entrants?) 
> will "give way free service."
> I don't know what this means actually. I'm assuming that you 
> mean that someone will (or should) provide some kind of 
> deeply (perhaps 100%) discounted service, the results of 
> which would be increased demand for native IPv6-based IDR -- 
> but I don't know who or what you have in mind specifically. 
> Would this be a recommendation that straddles one of the 
> other points above (e.g., for native IPv6-based mobile 
> access/access providers, or IPv6-based content hosting 
> businesses?), or something else entirely?

I know as I'm sure you do that we aren't going to see wide-spread
giving away of free service.  What I was discussing is that
the progression to adoption is going to specifically be a unique
problem for the US because of 2 factors, first the US sources a
lot of content on the Internet, second the US has much more
built-out IPv4 infrastructure.  You will see US content providers
dual-stacking early, BUT when most end-users and content-providers
are dual-stacked, it will be the end-users wanting to single-stack
to IPv6 and the content providers holding things back - and the
content providers in the US will be the worst about becoming
IPv6-only because of their heavy investment in IPv4 infrastructure.

Obviously, ISP's and networks (like the cell network) that provide
connectivity to end-users are the biggest consumers of IP addresses.
It is quite logical that they would be the most interested in
getting rid of IPv4 - once all of the content providers on the
Internet have dual-stacked.  So once the content providers have
done it, the retail ISP's will feel comfortable levying surcharges
and suchlike to their customers to push them into IPv6-only.  But of
course, not all customers will do this and as long as some are
foot-dragging, the content providers will stay dual-stacked.


> Sorry if this re-parsing seems excessively literal; I'm 
> trying to enumerate specific factors that should be included 
> in a formal model for simulating the evolving probability of 
> establishing/sustaining native IPv6-based inter-domain 
> routing services.
> Thanks again,
> TV
> On May 15, 2009, at 6:41 PM, Ted Mittelstaedt wrote:
> >
> > Here is how I envision this thing happening.  You can 
> adjust the dates 
> > as you see fit.
> >
> > Sometime in 2013 IANA and rest of the RIR's run out of 
> "virgin IPv4".  
> > The general news media starts making a big deal out of it.  
> Many ISP's 
> > who had no clue will get clueful.  ARIN will embark on 
> > "low-hanging-fruit"
> > reclamation projects to attempt to pull back abandoned IPv4.  This 
> > will satisfy the smaller requestors.  The larger requestors (eg. 
> > Comcast) will embark on "internal IPv4 reclamation" projects.  The 
> > Internet will still generally run on IPv4.  Cell phone 
> providers will 
> > begin heavily deploying
> > IPv6 along with IPv6-IPv4 proxy systems
> >
> > Between the years of 2013 and 2015 the RIR's will gradually exhaust 
> > "low-hanging-fruit" IPv4 reclamation projects and a paid, transfer 
> > market in IPv4 will arise.  By 2017 this transfer market will be in 
> > full swing and it will be regarded as customary for "aspiring IDR 
> > participants"
> > to
> > purchase IPv4 blocks.  Large orgs will be in full swing 
> with internal 
> > IPv4 reclamation projects, both for their own needs and to sell to 
> > newcomers.
> > Content providers who run websites will be heavily pressured to 
> > connect to the IPv6 network.
> >
> > By 2020 we will see the end of the transfer market as price 
> increases 
> > push block prices to the point that they will never be able to 
> > generate a return on investment.  By now, all orgs will be 
> deploying 
> > IPv6 as SOP for "native" IP connectivity, along with 
> RFC1918 addresses 
> > for IPv4 connectivity, or extra-cost public IPv4.
> > Small orgs will likely be doing the same as well as heavy 
> > experimentation and use of web and other proxies to get 
> IPv6 assigned 
> > address customers to
> > IPv4 providers on the Internet.  Large orgs will have 
> forced all major 
> > content providers to offer content via IPv6 in addition to IPv4
> >
> > From 2020 to 2025 will be the IPv4 end game.  The paid 
> transfer market 
> > will have collapsed and no large requestors will be asking for IPv4 
> > from the RIR's.  We will have the appearance of IPv6-only content 
> > providers, and this trend will accelerate.  New IDR participants 
> > likely will be only requesting nominal amounts of IPv4 to assist in 
> > accessing the remaining content providers who haven't switched.
> >
> > By 2025 everyone offering a website on the Internet will be 
> IPv6.  An 
> > increasing number of Internet customers will decline even the free
> > RFC1918 IPv4 addressing.  Gaps and issues with route propagation of 
> > the
> > IPv4 BGP table will become increasingly regular.
> >
> > By 2030 IPv4 will have been reduced to a handful of sites still 
> > advertising it mainly because they have not bothered to 
> clean up their 
> > internal networks.
> > A majority of transit AS's will be actively filtering IPv4 
> > advertisements.
> > IPv4 will essentially be dead.
> >
> >
> > So, in answer to your question of:
> >
> > "...what individual actions
> > -- either undertaken directly by the aspiring IDR 
> participant, or by 
> > the operators of other, established routing domains --  
> could directly 
> > or indirectly impact the probability, timing, and extent of this 
> > becoming generally possible?
> >
> > I would say:
> >
> >
> > Sign up more customers.  Get people on the Internet who are not 
> > currently on the Internet.  What ultimately pushes IPv6 is 
> a shortage 
> > of IPv4, and the only thing that creates this is 
> consumption of IPv4.
> >
> > This is likely why the United States is going to have the 
> worst time 
> > to switch to IPv6.  We have likely reached saturation for 
> the Internet 
> > market, and thus everyone in the country already has an 
> IPv4 address.  
> > What is going to drive the US to IPv6 is the rest of the world 
> > offering content on IPv6 and US customers wanting to get at it.  
> > Unfortunately, the US is the largest supplier of content in 
> the world, 
> > as George Carlin used to say:
> >
> > "America's most profitable business is still the manufacture, 
> > packaging, distribution and marketing of bullshit"
> > ---George Carlin: "You Are All Diseased (1999)"
> >
> > Ted
> >
> >> -----Original Message-----
> >> From: ipv6-ops-bounces+tedm=ipinc.net at lists.cluenet.de
> >> [mailto:ipv6-ops-bounces+tedm=ipinc.net at lists.cluenet.de] 
> On Behalf 
> >> Of Tom Vest
> >> Sent: Friday, May 15, 2009 11:15 AM
> >> To: IPv6 Ops list
> >> Subject: Factors,actions influencing the possibility/timing of IDR 
> >> for IPv6-basedrouting domains?
> >>
> >> Hi all,
> >>
> >> Would be very grateful for assistance from list members....
> >> apologies for any duplication...
> >>
> >> I'm trying to compile a list of environmental factors or 
> actions that 
> >> might affect the probability and timing of direct participation in 
> >> interdomain routing becoming practical, specifically for routing 
> >> domains that start out with some IPv6 but without even one 
> publicly 
> >> routable IPv4 address.
> >>
> >> The question I'd like help with is: what individual actions
> >> -- either undertaken directly by the aspiring IDR 
> participant, or by 
> >> the operators of other, established routing domains --  could 
> >> directly or indirectly impact the probability, timing, and 
> extent of 
> >> this becoming generally possible?
> >>
> >> Please note (if it wasn't already obvious) that I'm *not* talking 
> >> about the possibility of becoming a pure/direct customer 
> of another 
> >> routing service provider, nor am I asking/making 
> presumptions about 
> >> IDR-related activities that are not commonplace, much less 
> >> "guaranteed" for routing service providers in general (ubiquitous 
> >> settlement-free peering, etc.). The basic idea is that there is a 
> >> range of "normal" or "conventional" activities that operators of 
> >> routing domains may choose to participate in today (e.g., 
> exchanging 
> >> traffic, indirectly or directly, potentially with most if 
> all other 
> >> routing domains; participating as a third party in traffic 
> exchange 
> >> between other routing domains -- aka providing transit; pursuing, 
> >> accepting, and/or rejecting direct and indirect traffic exchange 
> >> relationships; influencing traffic flows across those 
> >> interconnections, etc.), none of which would be possible 
> today for an 
> >> operator of a pure IPv6-based (or any post-IPv4 runout) routing 
> >> domain. The question is: what's going to change that?
> >>
> >> My current provisional list of things that might 
> incrementally change 
> >> that includes:
> >>
> >> 1. New entrant(s) obtain an independently routable 
> quantity of IPv4 
> >> from someone/somewhere, which can be used to mediate 
> traffic exchange 
> >> between internal IPv6-based resources and external 
> IPv4-based network 
> >> elements.
> >> 2. Existing, IPv4-based routing domains offer native IPv6-based IP 
> >> transit that is physically accessible by the aspiring new 
> entrant(s).
> >> 3. Existing, Pv4-based routing service providers begin to 
> accommodate 
> >> incremental growth of existing and new customers using 
> IPv6 *in a way 
> >> that makes those new elements transparently reachable via native 
> >> IPv6- based IDR.* 4.
> >> Existing, Pv4-based routing domains make some or all of their 
> >> current, public-facing resources reachable via native 
> IPv6-based IDR.
> >>
> >> What else should be on the list? Additional comments, 
> questions, or 
> >> suggestions would be greatly appreciated!
> >>
> >> Thanks in advance,
> >>
> >> TV
> >>
> >>
> >>
> >>
> >>
> >>
> >>
> >>
> >>
> >>
> >>
> >>
> >

More information about the ipv6-ops mailing list