Re: Metadata over IPv6

[Brian Haley <[e-mail]>]

On 12/17/19 2:38 PM, Templin (US), Fred L wrote:
> Hi Brian,
> 
>> -----Original Message-----
>> From: Brian Haley [mailto:[e-mail]]
>> Sent: Tuesday, December 17, 2019 11:11 AM
>> To: Templin (US), Fred L <[e-mail]>
>> Cc: [e-mail]
>> Subject: Re: Metadata over IPv6
>>
>> Hi Fred,
>>
>> Thanks for the response, I do have some questions.
>>
>> On 12/17/19 12:32 PM, Templin (US), Fred L wrote:
>>>> For embedded IPv4 addresses/prefixes, however, we would use a modified
>>>> IPv4-compatible encapsulation as follows:
>>>>
>>>>     fe80::ffff:a9fe:a9fe
>>>
>>> I'm sorry; I intended to say "modified IPv4-mapped encapsulation". "IPv4-compatible"
>>> is deprecated by RFC 4291 and would contain all-zero's instead of 0x0000ffff in bits 64-95.
>>> AERO uses IPv4-mapped for embedded IPv4 addresses and all-zero's for administratively
>>> assigned AERO addresses.
>>
>> So in my case I don't think we want to define an IPv4-mapped
>> encapsulation, just a new on-link anycast address that something on the
>> link will answer to, it doesn't have to be the router since there might
>> not actually be one.
> 
> There is already a link-local Subnet Router Anycast address - it is fe80::.

Ack.  In our case there might not be a router, a network can be 
completely isolated with only a DHCP server attached, which can also be 
used as a proxy to metadata - it would return a route to the fe80::$META 
address via itself in that case.

>>> Also, fe80::ffff:a9fe:a9fe can be written as fe80::ffff:169.254.169.254 - the two address
>>> representations are equivalent.
>>
>> I wouldn't mind using the AERO format, but reading the draft you listed
>> it doesn't look like fe80::ffff/96 is being reserved by IANA?
> 
> fe80::ffff/96 is a format that has significance on AERO links. AERO is an IPv6-over-foo
> document that specifies the construction of link-local addresses used on the link. The
> format may be useful for other applications and/or link types besides just AERO,
> though, so it would probably make sense to adopt a consistent format. RFC7421
> cites the AERO format.

Ah, Ok.  So does everything on the link use the same address format?  We 
are dealing with pre-existing deployments where we can't change that.  I 
do like the fact that AERO basically has the format we're looking for, 
fe80::ffff:169.254.169.254, which make it obvious this is the metadata 
service IP.

-Brian


>> Just that
>> "Relay, Server and Proxy AERO addresses are allocated from the range
>> fe80::/96, and MUST be managed for uniqueness."
> 
> You are referring to something different here that does not apply to the use case
> of embedding an IPv4 address in an IPv6 address - so, this part of the AERO spec
> is out of scope for what we are discussing here.
> 
> Thanks - Fred
> 
>> since we do control
>> the MAC we could do this I suppose as we want this link-local on the
>> proxy, but what we're doing isn't really akin to AERO (IMO).
>>
>> Thanks,
>>
>> -Brian