[6bone] Address management transfer proposal

[Jeroen Massar]

'Anand Kumria' [mailto:wildfire@progsoc.uts.edu.au] wrote:

> On Tue, Dec 31, 2002 at 04:06:08PM +0100, Jeroen Massar wrote:

<SNIP>

> > Who/what says/defines that suddenly one has to 'pay' for 
> 6bone space?
> 
> It was part of the proposal, see 
> <URL: 
> http://www.apnic.net/mailing-lists/sig-ipv6/archive/2002/08/ms
> g00000.html>
> 
> 3.1 parts (c) and (d); note I didn't say "suddenly" but it will turn
> 6bone into a service which requires payment.

Quote from 3.1d:
8<-------------------
fees will be waived for 6bone address services provided by RIRs to
6bone members (but not for other services 6bone members may require),
until
1 year after this agreement starts.  After this time each RIR may charge
an
administration fee to cover each allocation made.  This fee simply
covers
registration and maintenance, rather than the full allocation process
for
standard RIR members.
----------------->8

Waived -> No Payment for the first year.
After that only 'setup' costs.

And that for something that is just like 'the real thing' (RIR space)
and for a less cost. Organisations (not individuals) that have a good
enough backbone won't really worry about a small fee like that.

> Specifically from the APNIC 2001 annual report; they make USD$3620K 
> revenue. At least USD$581K at directly attributable to IPv4 address
> space delegation.

Those numbers have nothing to do with 6bone, also those are membership
fees and other numbers coming from different sources.

> Make no mistake that IPv6 represents a revenue threat to all RIRs. 

"Revenue", the RIR's are *NOT* in it for making money but for providing
a *SERVICE* to their members. Again, 'money' has nothing to do with
6bone.

> > > Keeping existing system:
> > > 
> > > Pros:
> > > 	- no changes
> > > 
> > > Cons:
> > > 	- may induce volunteer burnout
> > > 	- no e.f.f.3.ip6.arpa delegation
> > > 	- doesn't excourage migration to production IPv6 addresses
> > 
> > This would eNcourage migration to production IPv6 as the 6bone
> > would be 'harmed' by the nonexistent ip6.arpa delegation.
> 
> Perhaps, my current feeling is that there are more end-users 
> using 6bone address space than IPv6 production space. I have no data
to 
> back up that feeling though.

Are those users 'testing' or using it 'production' ?
If they are using the second they should be using RIR space anyways.
And again, <cut from other messy>:
8<---------
ripe:  140
6bone: 134
apnic:  93
arin:   42

Hmm 134 6bone TLA's and only 140+93+42 = 275 RIR TLA's...
--------->8
275 RIR TLA's is the double of 6bone and you can expect
them to be used too as people are paying for the
*REGISTRATION* and companies DO need to produce revenue.

> Those end-users of 6bone will eventually start to demand RIR IPv6
> addresses.

Indeed and that is *GOOD* as they will harrash the marketing
departments of their uplink and then it will become production IPv6.

> > Also many people are moving on to RIR space, at least I think
> > that is what you mean that with 'production' IPv6.
> 
> Yes, I do; I'll use that term instead (IPv6 RIR space) instead.
> 
> > > Some points I noted, my comments are under them:
> > > 
> > > + RIRs have no incentive to pull-in IPv4 and hand-out IPv6; I 
> > > think this will guarentee even slower IPv6 rollout
> > 
> > I think you are quite wrong here; See
> > http://www.ripe.net/ipv6/v6allocs.html
> 
> 404'd

http://www.ripe.net/ipv6/ipv6allocs.html

Never type URL's from the head...

> > or http://www.sixxs.net/tools/grh/
> 
> A very useful tool!

And it will be showing bogon's quite soon too, be warned abusers ;)

<SNIP>

> > Hmm 134 6bone TLA's and only 140+93+42 = 275 RIR TLA's... 
> not enough? :)
> 
> Interesting; and yet end-user IPv6 is still hard to find the 
> world over.

What do you mean with "end-user IPv6" ?
You probably mean "IPv6 where people pay for" and there is enough
of that in the APNIC region, you can even get IPv6 dedicated lines ;)

> > Only the ARIN region is quite behind, but the others are growing
> > rapidly.
> 
> Actually I'd say that APNIC is also.

APNIC region is much smaller then the RIPE region, covering europe,
most part of africa and russia.
Also an ISP only needs one /32 to fullfill most of it's needs
currently an ISP will most of the time have multiple IPv4 /19's.
These are IPv6 /32's with which you can address 2^(128-32) IP's.
Which should be enough for 90% of the biggest ISP's around the globe.

> > > It'd be nice if RIRs started to penalise their large 
> transit/backbone
> > > operator for requesting IPv4 space but not have/using 
> IPv6 space. It'd
> > > also be nice if pigs had wings too.
> > 
> > ISP's/Transit providers are BUSINESSES. They have to earn money.
> 
> Yes, so you'd imagine that: lower fees for RIR IPv6 coupled with
> additional costs for RIR IPv4 would encourage migration.

No it won't as there still is no DEMAND for IPv6.
It's the same as what philips tried to do here in .nl:
Make the "widescreen" (16:9 ratio) TV's cheap and the
"standard" (4:3) more expensive, but that didn't make
TV-stations broadcast any 16:9 programs. The only use is
watching DVD's (imho 1m^2 screens are still to small though ;)

> Most people on this list are technical and neglect to take 
> into account that beancounters have a lot of sway also; demonstrate
that 
> IPv6 is less costly and perhaps more organisations will commence take
up.
> 
> Most beancounters that I've met are very happy if they can lower their
> OPEX [1] (operational expenditure) by a small amount of CAPEX 
> [1] (captial expenditure).

Those beancounters also say that they are doing "badly" when they
didn't make the money they expected to be making even though
everything is up up up and they had doubled their profits etc.
They are mostly in it for the stockmarket, not for the technical savvy.

> > > + having the delegation, as far as I can tell, depend on migrating
> > > address management to the RIRs will only slow down IPv6 adoption.
> > > 
> > > I help out on various irc channels people setup their tunnels; the
> > > hardest thing for most of them is to get reverse DNS gonig. When
they
> > > discover that they have to do it twice (for ip6.int and ip6.arpa)
most
> > > of them don't bother -- even if the work isn't much.
> > 
> > Give those people clue first and probably the only reason why they
> > want reverse is to 'look cool on irc'. 
> 
> Perhaps; personally I've not encountered that. But then I lead a
> sheltered irc existence.
> 
> > That's not a reason to do IPv6.
> 
> What is? What you mean is that you don't think it is a valid reason to
> deploy IPv6; I'm sure a clever ISP could start to make money form it
> though. Vanity can earn $$$ dollars.

Vanity IRC hostnames is not a reason to do IPv6.
Even though for most people it is apparently.
And Vanity earns loads of money, seeing a 'free email' company having
1500 domains (<dom>.nl that is), good money for SIDN :)

> > > Having the RIRs not delegate e.f.f.3.ip6.arpa means that 
> > > support will be
> > > in place in a lot of OS for ip6.int for many years to come.
> > 
> > 6bone is a testbed it ain't that critical.
> 
> True; however at the moment all software has to support both 
> ip6.int and ip6.arpa;

Is that *that* hard, it has been that way for almost two years now.
Most (if not all) software understands it.

<SNIP>

> > > + 6bone may not fall under the IETF IPv6 AD 
> responsiblity; hence the
> > > desire to more it "somewhere approriate".
> > 
> > Could you translate that?
> 
> Apparently, from the archive I read, the genesis of this transfer was
> because the IETF IPv6 Area Director (Randy Bush?) was/is in 
> the process of winding up the ipngwg; the 6bone project did not fit
into the terms
> of reference of the next one. Thus the desire to find a another home.
> 
> [[ Apologies if I've got this all wrong, clarifications welcome ]]

That is indeed correct as ngtrans has become ipv6ops, because
IPv6 is 'operational' and that 'testing' has been concluded.
But as 6bone is quite broadly used they are seeking a nice way
of restricting it and forcing it into the RIR's. Which is a good thing.
Current 6bone TLA holders will end their actual 'productional' use of
the 6bone and move to RIR space also wrapping up the endless fullmeshed
tunneling. 6bone TLA holders that actually are really using their
space for testing will continue to do so. But it should all go away.

> > Convince them that they should; they will probably have one big and
> > fairly good argument: Pay us.
> 
> Actually the AS7474 people said they hadn't seen much demand; so I've
> setup a tunnel broker and lo' 50% of the people taking up the service
> are within AS747.

Show the stats to AS7474 then, possibly letting them run a tunnelbroker
for their endusers.

This is actually why Pim and me have setup SixXS:
ISP wants to give their endusers IPv6 connectivity, but they don't have
the proper hardware in place to get IPv6 to their endusers.
Thus they:
 - Get a RIR delegation.
 - Build a tunnelbox (simple PC or similar)
 - Install the SixXS software.

Et tada, they can provide their users (or a specific set of prefixes
etc)
with IPv6 without having to go through the 'problem' of buying/upgrading
their existing infrastructure. Some places (*DSL etc) simply cannot be
upgraded easily or if they can it is not in the budget of that ISP.
Every ISP should be able to spare a simple PC with a NIC though.

> From what I understand from my second hand hearing of things; they are
> now looking at their suppliers to see which support IPv6. 
> Unfortunately it is a large organisation so those kinds of
> (re)evaluations can take time.

Politics... yuck ;)

> > > I feel that transferring the address allocation to RIRs 
> (merely for
> > > the DNS delegation) will lead to the hastened end of the 
> 6bone. We all
> > > realise the 6bone will, and must, go but I think it is too 
> > > early at the moment.
> > 
> > I read this as "because I don't have ip6.arpa my irc doesn't work".
> 
> Read it as "because ip6.arpa isn't delegated growth of IPv6 
> demand (from end users) is slowed".

Reverse DNS doesn't impose any problems on HTTP,SSH,Mail, actually
*any* other protocol except IRC (and maybe some log programs).
But still, most resolvers understand both ip6.int. and ip6.arpa.

> [1]: Widely used terms within Australia, not sure about the 
> rest of the world.

Not something one will find in non-native-english speaking at least ;)

Greets,
 Jeroen