[arin-ppml] IPv4 is depleted today

[David Farmer]

On 29 Aug 2008 Ted Mittelstaedt wrote:

> On 29 Aug 2008 David Farmer wrote:
>  
> > On 29 Aug 2008 Ted Mittelstaedt wrote:
> > 
> > > There are a couple of phrases in this post I found most
> > > humorous:
> > > 
> > > "...return our current IPv4 address space it is to valuable 
> > > internally...."
> > > 
> > > "...will be necessary to extend the life of
> > >  IPv4 long enough for us to get IPv6 ramped up..."
> > 
> > Glad to be comic relief for you, but;
> > 
> > > What our esteemed colleague David is saying is in essence
> > > he wants a transfer policy because he has a lot of IPv4
> > > and he (understandably) wants to continue to use it.
> > > He labels IPv6 not ready, why not?  He already has
> > > plenty of IPv4, he doesn't need IPv6.  He has no incentive
> > > to deploy IPv6, really.
> > 
> > Wrong, we have plenty of incentive to deploy IPv6, we been 
> > working on it, 
> 
> Can you walk into your University Presidents office and
> tell him with a straight face that unless they pony up a
> quarter million bucks over the next 3 years for IPv6 deployment,
> that the college will lose connectivity to the Internet?

I don't have to; IPv6 was built into our requirements for our last tech refresh, 
4 years ago, and that wasn't a quarter million it was 16 million, and included 
10G backbone and 10/100/1000 for now 70,000 ports.

> That is my point.  Of course you could not.  

Probably not as a separate project, but if your not doing a tech refresh in the 
next 2 years or so why didn't you include IPv6 in your last tech refresh?  The 
writing was on the wall.

> If you did there
> would be an investigation by the beancounters who would find
> out that the real answer is "well, maybe we need to do this
> but not right at this moment" and the President would say "I
> understand where your coming from but there's more important
> things to spend money on right now" and that would, as they
> say, be that.

No, that wouldn't be the President of the University that would be the CIO, 
and I have a plak on my wall thanking me for a three year effort, of which 
making sure all the new equipment has basic IPv6 capabilities was one little 
part.

> THAT is what I am talking about when I say you have no real
> incentive.  Not you personally, your organization.  It does not.
> Have any real incentive.  And it will not until the rest of
> the Internet does, in fact, switch to IPv6

That is where you are wrong, you might be right for other organizations but 
we were doing IPv4 in 1985 or so.  What was our incentive to do IPv4 then?  
We have the same incentives to do IPv6 now. 

> You must have heard of the phrase "Mexican Standoff" I am sure.
> Same issue.
> 
> > have you?  We just have bigger incentives not to break IPv4, 
> > yet, and there 
> > are other projects, with higher priorities too. 
> 
> So you will do very little until the loss of access to IPv6
> makes conversion to IPv6 into the top-dog incentive, more
> of an incentive than breaking IPv4, or other projects with
> "higher priorities"

It will be done long before it needs to be a top priority, if I wanted to just 
throw the switch I could do it right now.  But, we aren't just going to throw the 
switch. Did you miss the part in my original email "I hope to enable it (IPv6)
on our whole backbone and make it available anywhere anyone wants it yet
this year or early next, then in 2010 or 2011 turn it on by default
everywhere on campus."

> This is visionary management?  

Well, I think we do pretty good, if your not ready for IPv6 then it is you who 
lacks visionary management.

> Aren't you setting the priorities on your own projects? 

No, that is what CIOs, Directors, and Managers are for.  We are deploying 
about 3000 new 802.11n APs as we speak, students don't care about IPv6 
yet, they sure do about wireless though.  We just completed upgrades for 
10G WAN connectivity, including lighting 1500 miles of dark fiber, we share 
a 100G to Chicago and are already upgrading to 200G.  IPv6 is not a top 
priority, but it high enough on the list that work does get done on it.

> > But, we are 
> > making deliberate, 
> > all be it slow progress, are you?  Four years ago we had IPv6 as an
> > equipment requirement for our new campus network RFP, we have been
> > putting our money where our mouth is for IPv6 for a while.
> > 
> > But, IPv6 ain't ready for grandma today, with a lot of work 
> > it will be by IPv4 
> > exhaustion.
> > 
> 
> And when will this work be done?  It seems the incentive is into
> putting work into finding out how to avoid doing the work to switch to
> IPv6, rather than actually switching.

Not for us.  We've been working on it for several years now.

> Name 5 reasons that have nothing to do with money that you
> could not switch to IPv6 right this moment.  I am sure there must be
> some.  THOSE are the REAL reasons that are blocking IPv6 deployment. 
> All other reasons, including the monetary "other projects are higher
> priority" are bogus nonsense.

1. The risk of getting IPv6 labelled as that stuff that breaks the network
2. Testing, Testing, more Testing, and Change Control
3. Lack of feature parity between IPv4 and IPv6 on the same platform
4. Stupid vendor games, making IPv6 a special feature train 
5. Having to argue with people like you that IPv6 is what needs to happen

See this for the detail of a particular instance of #2 and #3;

http://events.internet2.edu/2008/jt-
hawaii/sessionDetails.cfm?session=3638&event=278

Also, #2, #3, and #4 have money components, but thats not the driver for 
them.

#5 alternates between a big a head ache and a lot of fun depending on how 
I feel on a particular day. :)

> When people talk about IPv6 not being ready, THOSE are the issues that
> they should be talking about.  But they don't - mostly, they talk
> about how expensive it's going to be to switch everything. That's
> making excuses.

Not us, look at the link above.

> In our case the reason is simple - our major upstream promised native
> IPv6 this fall, last year.  A few months ago they reniged.  There will
> be consequences when their contract comes up at the end of the year. 

Good, you have to put your money where your mouth is.

> And incidentally they are a GSA contractor and so they obviously are
> reniging on the US government requirement for IPv6 too.  I cannot
> understand how this is allowed to be so without consequences - but
> thankfully our owner has avoided government jobs when possible so we
> do not have, as they might say, an intimate knowledge of how
> government suppliers are allowed to get away with breaking their
> promises.  Our suppliers aren't.

I won't go there except to say, the Government "requirement" was mostly 
smoke.  I have a commercial IP provider (GX) that does IPv6 and IPv6 was 
part of the reason we selected them.  I also have Internet2 connectivity.  
You could do tunnels to get started, I wouldn't use tunnel for production, but 
you can get your feet wet.

> > > He also won't give his network's IPv4 up to "sell" through 
> > a transfer 
> > > policy, or course not.  He is just expecting every other 
> > legacy holder 
> > > out there to sell off their IPv4.
> > 
> > Wrong, if the University of Minnesota were to stop using it's 
> > IPv4 address 
> > space we will return it to ARIN, and not sell it.  We have in 
> > the past, we 
> > returned a C to IANA when we got our original B, and we may return a
> > couple other swamp C's in the future as a matter of principle, but
> > that really won't make any difference to anything.  Nor, would
> > returning all of our address space, it might delay IANA exhaustion
> > by a whole day or two. 
> > 
> > I'm just trying to recognize that it will be easier and more 
> > effective to give 
> > people an incentive to return address space, then to only ask 
> > them return it, 
> > or to try to force them to return it.    
> > 
> 
> Timing is everything.  Giving back a large swath of IPv6 pre-IPv6
> runout matters.  Giving it back soon after runout matters.  Giving it
> back years later, when IPv6 is well on it's way, it nothing more than
> a photo-op.

I think you mean "Giving back a large swath of IPv4 pre-IPv4 run-out 
matters."  At the current burn-rate giving back anything less than a major 
portion of a /8 is really only a photo-op.  That might means a month or two 
longer at current burn-rates.  After IPv4 run-out giving back any addresses 
probably wouldn't even make a historical footnote, let alone a photo-op.

> > > The situation reminds me when my city, Portland OR, put
> > > in Light Rail.  Everyone driving cars in the city strongly 
> > supported 
> > > light rail.  They all wanted it because the figured that "the
> > > OTHER guy" would stop driving his car, and take light rail, and 
> > get the heck 
> > > off the freeway so that THEY could drive an uncongested freeway.
> > > 
> > > Needless to say, the freeway congestion did NOT go away.
> > > 
> > > I would ask our colleague that if he feels that it would
> > > be too much trouble for him to split his legacy block
> > > to be able to sell off part of it under a liberalized 
> > transfer policy, 
> > > doesn't he think that everyone else with a legacy block is going
> > > to feel the same way?  Why is he special?  And if every other
> > > legacy holder feels this way, then where exactly are these IPv4 
> > blocks going 
> > > to come from that will be sold through a liberalized transfer
> > > policy, that will "extend the life of IPv4" as he puts it.
> > > 
> > > In short, he meticulously explains why a liberalized 
> > transfer policy 
> > > would do absolutely nothing to help him return his unused 
> > IPv4 space 
> > > to the free pool, then proceeds to claim that we need a
> > > liberalized transfer policy to enable more IPv4 to be returned to
> > > the 
> > free pool!!!
> > > 
> > > I can't think of a more damming example of why a 
> > liberalized transfer 
> > > policy would be about as useful as teats on a boar.
> > 
> > You are right, we're not special, a liberalized transfer 
> > policy will probably not 
> > free up much space.  That was one of my points, that I 
> > obviously didn't 
> > hammer hard enough.  
> > 
> > But, it might free up a little, and we're going to need every 
> > hair-brained Wile 
> > E. Coyote idea we can come up with to get to keep IPv4 going 
> > long enough 
> > to get everyone on to IPv6.
> 
> Correction WE are going to need it, YOU are not.  Because as you
> say, you already have enough IPv4 to last you.

Because of Metcalfe's Law, I = YOU = WE = THEM, I lose value if your not 
connected. 

> > Because unlike Wile E. Coyote, 
> > if the Internet 
> > falls off the cliff we not going to get up and walk away.
> > 
> 
> No, what will happen is the moneymen in the various companies
> that have billions invested in the Internet will apply wads of
> cash to the problem and voila, all IPv6 deployment problems
> will vanish.

Right, you obviously don't understand the combination of Internet Scale and 
Logistics, there are problems that money can't solve, and that only time can. 
Wile E. Coyote seems to have an infinite credit line with ACME and he still 
can't catch the Road Runner. :)

> Right now all they are doing is throwing engineers at the problem and
> letting them argue amongst themselves as to the best solutions.
> Because, they really don't care right now because there's still IPv4
> available.  They figure once runout actually happens, the engineers
> will have finished arguing and then they will start paying for the
> deployment.

That's to late, the Internet goes splat, and unlike Wile E. Coyote we don't get 
up a walk away.

> > I'll put it back on you, are you sure we can get IPv6 fully 
> > deployed before 
> > IPv4 crashes into the wall, I'm not.  We (UMN) will be there 
> > but we've been 
> > working on it for a long time.  I don't think everyone will be.
> > 
> 
> Then, those who truly are unable to deploy it will say goodby
> to the Internet and go back to growing vegetables, or mailing
> letters via US mail or whatever they were doing before the
> Internet came along.  While the rest of the world leaves them
> behind.

Again Metcalfe's Law, if they go away we lose a lot too.

> When the horse-drawn carriage came out, a lot of people who were
> farmers ended up being blacksmiths, and carriage makers, and
> such.  Then when the automobile came out, not every buggy whip
> maker made the transition to making cars.  Some kept making
> buggy whips until they died - and likely, they were happy
> doing it.  They had just opted out of the rat race.
> 
> We cannot fundamentally change the Internet to accommodate the
> buggy whip makers.  Right now we all are making buggy whips.
> But, before most of us retire, we likely will be making cars.

Then why do you seem to be telling people to keep making buggy whips?

> > If we have a transfer policy that might free up a little 
> > space, it is not going to 
> > save us in the long term, nor will NAT, only IPv6 can do 
> > that.  But it could 
> > help build a bridge,
> 
> No, all it will do is provide incentive to the beancounters
> who don't know squat about anything, to NOT spend money for
> a while longer.  It's a false hope, a mirage.
> 
> > and if IPv4 addresses went for $1000 an 
> > address it sure 
> > would make a bigger incentive to move to IPv6 and hopefully 
> > push IPv6 up 
> > the priority list for IT departments everywhere.  So if only 
> > for that reason, it 
> > might help to have a transfer policy, to give that extra little
> > push.
> >
> 
> IT departments who would only take action if this happened are what we
> call reactionary.  In the US, unless a business is in a monopoly,
> (like Microsoft) being reactionary just pushes you down the whirlpool
> of inaction until you go out of business.
> 
> > I also think the longer we debate a transfer policy, the more 
> > the message 
> > that IPv6 is the only true hope get diluted. 
> > 
> 
> You are right.  We should just kill the transfer policy idea and
> be done with it.

I could live with that if it could only just happen.

You might subscribe to my alternate proposal then, lets relax the usage 
requirement for IPv4 and kick up the burn-rate, make it run out faster.  Then 
your moneymen come to the rescue.  The only problem with this, that's a big 
gamble.  If we get it wrong, Wile E. Coyote goes splat and maybe even hard 
enough that he doesn't get up.

What do you think?

> Ted







=======================================================
David Farmer				     Email:	farmer at umn.edu
Office of Information Technology
Networking & Telecomunication Services
University of Minnesota			     Phone:	612-626-0815
2218 University Ave SE			     Cell:		612-812-9952
Minneapolis, MN 55414-3029		     FAX:	612-626-1818
=======================================================