Re: When IPv6 ... if ever?

[Eliot Lear <lear () cisco com>]

You know, Sean, take a pill.

Here is what you wrote:

smd () clock org wrote:
> Also, nobody is willing to get shot!

Unless you're (a) a startup or (b) a VERY big company who can direct the
market.

> The deployment of IPv6 is going to be EXPENSIVE in terms of real opex
> and probably real capex as well, it IS going to be visible on the bottom
> line of every ISP on the planet, eroding whatever margins one has.

This is true.

> I can't see the deployment of IPv6 *ever* leading to any but the
> shortest-term revenue upside (if even that), therefore until the
> entire aggregate gross revenue of transit-providing ISPs up and
> down the entire food-chain is threatened, nobody will be deploying v6.

This is false.  In the end ISPs will be able to make a wash of it
through pricing structures.  First there are early adopters, and those
are here now.  That grows into a small group of networks.  Those are
likely to be here next year.  All it takes are a handful of large ISPs
to say, "I'm game", and it's amazing what the shape of the net looks
like.  Anyone who disputes this is disputing history, since this is
precisely what happened with previous improvements, the invention of the
FIXes, MAEs and BGP.

> The only alternative scenario I can think of is the deployment
> of IPv6 by a large provider who believes it can trigger a huge
> consolidation by pushing smaller ISPs it is competing with into an
> expensive deployment through sheer hype.

You're half right. All it takes are a large providers with a real
application.
 
> Everyone who wants cheap, sustainable Internet transit, with the
> continuation of plummeting prices combined with soaring available
> bandwidths.   Introducing a whole new protocol requiring massive global
> operational changes is going to push up consumer prices and stall
> on investment in available bandwidths.  There are only so many people
> and dollars out there, and one or the other is inevitably spread
> pretty thin in the current market.

I guess the major point your missing is that once a major provider goes
the others are going to realize that the Internet isn't going to shrink
in size, and so the cost of moving to v6 is only going to go up.  Hit
critical mass and the party really begins.

> The scariest thing to an IPv6-Lover is that an early deployment is
> not to anyone's advantage because untili there is real uptake by
> a sizable number of ISPs, the exact changes required on the dynamic
> routing side are simply not obvious, although the fact that the two
> protocols will run ships in the night is (e.g. www.microsoft.com
> works just fine with IPv6 but you see a blackhole with IPv4.
> ftp.cdrom.com's IPv6 path is much slower and lossier from where
> your customer is than ftp.cdrom.com's IPv4 path.  have fun finding
> and fixing the square of the number of problems you observe now, kids!)

I don't see this happening.  I see a lot of interest in running two
separate networks, where they might get merged later.

> In other words, it's all risk and absolutely no reward, and until
> it really honestly IS impossible to do hacks around the IPv4 shortage,
> nobody will deploy IPv6.

The other downside you fail to mention is the growth of customers.  It's
still positive, surprise.  And when is the problem more manageable?  A
further downside is the cost of IP address administration that continues
to climb.  If the assigning authorities can allocate out larger blocks
the cost of remaining at IPv4 becomes far more obvious.

> Who will take the chance of a huge investment in managing IPv6 deployment,
> when it is not a given that IPv6 really will be the header networks will
> use after IPv4?   We're talking about stranded assets being the only thing
> one gets for the money...

As it stands today there are 0 alternatives that are being seriously
considered by all parties, so while it's not a sure bet it is a good bet
to place some amount of time into and charge a premium for early
adopters (remember those)?  What do the early adopters get?  A chance to
reduce the number of times they will need to renumber, a task which
today is still quite expensive, even with tools such as DHCP.  Bill
Manning's dream is still far from reality.

Also, while I believe that routing is a very serious scaling issue, the
last time I looked, the growth rate in address allocation was picking
up.  This Is Bad <tm>.

This is not to say that I believe IPv6 to be the cat's meow.  I think
the proponents' marketing has been nearly as bad as your
anti-marketing.  It would be worse except as of late they seemed to have
toned down where you haven't.

I think the one area that will give people pause will be the header
size.  To me that's serious overhead for the often bespoke "killer app."

Eliot