[ppml] IPv6 getting real: was Policy Proposal: IPv4 TransferPolicy Proposal

[Kevin Oberman]

> Date: Sat, 16 Feb 2008 15:27:44 -0800
> From: k claffy <kc at caida.org>
> 
> On Fri, Feb 15, 2008 at 10:16:59PM -0800, Kevin Oberman wrote:
>   
>   Unlike the commercial world, the research and educational networks of
>   the world mostly provide full IPv6 capability. Some of us have been
>   providing production IPv6 for over half a decade.
> 
> kevin, this definition of "full" is as fuzzy as the vendors'.
> it took us (an network research group at SDSC/UCSD) 2 years to 
> get IPv6 connectivity to our prefix after we first requested it.  
> we hit 6 or 7 obstacles, including address space suballocation
> bureaucracy (weeks back and forth getting clarification on
> what kind of ipv6 addresses we should have, then months with ucsd
> trying to get a suballocation) to cisco router working ipv6
> image acquisition (months, several re-tries) to sysadmin
> learning curves ("ugh, what do we do about autoconf") 
> to fires in san diego burning down the house of the only SDSC
> network admin who had the magic combination of enable access
> and ipv6 clue on the magic router between us and v6 transit
> to Internet2 (months.).  all along we had the issue of what
> research grant should pay for the additional work, since NSF 
> certainly isn't interested in funding infrastructure, and we 
> had to take time away from funded research projects to work on it.  
> my hopes for having the academics blaze the ipv6 trail 
> were tempered by the reality of trying it ourselves.

I can't speak for UCSD, although they do have some pretty good
networking people. I've worked with some of them in relation to IPv6 in
the past and, the main issue was actually getting something to happen. I
don't know haw many queries I sent them (no, I'm not naming names) about
it, but it literally took years, about 2, even though I know that they
were already running IPv6 routinely in support of a remote microscopy
experiment. 

I suspect that it was just too low on their priority list.

ESnet treats IPv6 in exactly the same way as we do IPv4 and, when a
customer requests IPv6 connectivity, it is usually provided in less then
48 hours, providing their equipment/software supports it.

That said, we have had such little demand that I do worry that the
capability might atrophy. We get fairly regular queries about how to get
IPv6 connectivity, but follow-ups, let alone actual traffic often never
happen.

With the recent federal mandates for IPv6, I suspect that will change,
but I am not holding my breath. It does get frustrating. But I do not
think that the IPv6 support offered by ESnet or Internet2 is in any way
half-baked. Unfortunately, neither makes it to the end users that want
it (and I really believe that they exist).

>   The fact that IPv6 is available to most users at many major universities
>   in the US, Canada, and Europe should mean a fair amount of traffic. 
> 
> there may be some optimistic confluence of 'universities' 
> and 'users' here.  we have ipv6 to our prefix now, but we had 
> strong incentive because we want to do ipv6 topology mapping.
> i can't imagine why academics in general 'should' use ipv6.
> academics in general have no idea what ipv6 is, nor reason to learn.

I am sure that you are right. Most academics have little, if any
incentive to do IPv6. It's the students I would hope to see generating
some traffic. At least a little bit.

>   After all, it's in the core. You would think college students
>   would be trading MP3s or movies or something. (I've heard many rumors
>   that they have been known to do so over IPv4.)
> 
> you lost me there. if they're trading music and movies
> over ipv4, why should they use ipv6?  

Because many schools block the IPv4 file sharing on legal, moral, or
network survival grounds and, to this point, most don't even have the
capacity to do much about IPv6 because most of the tools for managing
file sharing don't work with IPv6. There is also the likelihood that the
MPAA and RIAA are not watching IPv6 activity. 

I have never been into such things, but, if I was and had IPv6
available, I think I'd go there.

>   Is there traffic? Not that I have seen. Is there demand? Not that I have
>   seen. Is there interest? At least a bit more than I typically see in
>   the commercial Internet, but not a whole lot.
> 
> well, i'm not sure how you'd see it since the Juniper routers
> that form the core of Internet2's backbone, while v6-capable, 
> are not v6-netflow-capable.  the flow export on those boxes 
> doesn't support v6, and there has been no other attempt to 
> characterize IPv6 traffic (tunneled or native) on I2. 
> (see slide 14 of joe's talk http://www.uoregon.edu/~joe/missing-half/)

I think we just hit the "half-baked" part. ESnet generally places IPv6
on separate VLANs or interfaces to let us measure IPv6 traffic. It's not
perfect, but it lets me say there is almost none.

> so in 2008 not only is there no evidence that U.S. academics are 
> using IPv6 applications on their backbone, there is also currently 
> no means to acquire evidence.   no trailblazing here.  and if
> I2 only had enough resources for tilting at one windmill this 
> decade: DNSSEC or IPv6, which should they pursue?
> 
>   Every major router vendor offers "full" IPv6 support. It's just that the
>   definition of "full" is a bit fuzzy. It often is synonymous with
>   "half-baked".
> 
> i'm still missing the leap of logic that connects 'half-baked' 
> to the expectation that academic networking staff on perpetually 
> tight budgets and already loaded down with more critical networking 
> issues than they can handle should be investing in IPv6.
> i'm not saying they shouldn't, but noone has made the case.

None. I am primarily talking to the wishful thinkers, mostly in the
commercial world, who keep mumbling "IPV6 is the answer" when asked
about almost any major issue the Internet is likely to see in the next 4
years. 

While IPv6 is important and I really, really believe it is coming, they
might as well be saying "42 is the answer".

> the R&E community will need exactly the same two things that the 
> rest of the world will need to support ipv6:  incentive and capital.

Once gain, you seem to think that my message was a commentary on R&E
use of IPv6. R&E is WAY ahead of the rest of the US Internet. That is
that the R&E community is doing something, if not much.

The real problems are in tools, support (e.g. lack of netflow either for
or on IPv6), and a near total lack of interest by those who could make a
real difference. (This means service providers and ISPs.)

Maybe I was not clear on my real points. I will admit that I was on a
bit of a rant and that does not always make for the best of clarity.
-- 
R. Kevin Oberman, Network Engineer
Energy Sciences Network (ESnet)
Ernest O. Lawrence Berkeley National Laboratory (Berkeley Lab)
E-mail: oberman at es.net			Phone: +1 510 486-8634
Key fingerprint:059B 2DDF 031C 9BA3 14A4  EADA 927D EBB3 987B 3751
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 224 bytes
Desc: not available
URL: <https://lists.arin.net/pipermail/arin-ppml/attachments/20080217/2e1f340a/attachment.sig>