Re: [v6ops] draft-ietf-v6ops-icp-guidance WGLC

[Brian E Carpenter <[e-mail]>]

On 07/08/2012 04:59, Cameron Byrne wrote:
> On Sun, Aug 5, 2012 at 10:29 PM, Fred Baker (fred) <[e-mail]> wrote:
>> This is to open a two week Working Group last Call on
>>
>> http://datatracker.ietf.org/doc/draft-ietf-v6ops-icp-guidance
>>   "IPv6 Guidance for Internet Content and Application Service Providers",
>>   Brian Carpenter, Sheng Jiang, 10-Jul-12
>>
>> Please read it now. We are interested in, among other things, technical commentary on the draft and the working group's perception on its usefulness to its target audience.
> 
> Useful.
> 
> Support.
> 
> Few technical comments here, few editorial comments sent direct to authors
> 
> Does this exist in the "real world"  ?
> 
> "An ICP that has multiple connections via multiple ISPs will have
>    multiple PA prefixes.  This results in multiple PA-based addresses
>    for the servers, or for load balancers if they are in use."

I know that there are instances of this, and it is certainly supposed to work.
Numerically, it's probably not important today, because early-adopter ICPs
are big enough to request a PI prefix.

> 
> It seems like NPTv6 is a much more modern approach that is much more
> likely to be deployed ... 

For content providers???

> or some method involving the load balancers
> doing a proxy / load balancing function for the new and legacy prefix.

That could be, but do we have any current practice to describe?

> 
> It would also be good to reference RA Guard (rfc 6105), ND Cache
> issues (RFC6583),  p2p /127 (RFC 6164)

Yes

> 
> It might also be prudent to mention that proxies are commonly
> implemented in load balancers.  Nearly all load balancers i know of
> have an Ipv6 to ipv4 proxy function (including Amazon ELB)

Yes, if you are doing that sort of LB. There are other approaches,
but that's a whole other topic.

> 
> 12.  Operations and Management -- might be worth noting that many
> large entities have not fully made IPv6 work just like IPv4 in OAM...
> and in many cases, some hashing or aliasing is used to represnt the
> 128 bit IPv6 address into the legacy IPv4 data base field.  This is
> not what people should do, but this is what many "IPv6 leaders" have
> done to make IPv6 work without boiling the ocean in terms of project
> costs and timelines.

Oh yuck! If we mention this, surely it should be in a negative way?

Thanks

   Brian