Re: [IPv6] I-D Action: draft-ietf-6man-rfc6724-update-04.txt

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

Mark,

On 28-Nov-23 19:16, Mark Smith wrote:
> Hi,
> 
> On Tue, 28 Nov 2023, 15:34 Brian E Carpenter, <[e-mail] <mailto:[e-mail]>> wrote:
> 
>     On 28-Nov-23 12:57, Mark Smith wrote:
> 
>      > I don't entirely understand why ULAs ended up with a global scope,
> 
>     Because:
> 
>     a) the site-local virtual experiment showed that "site scope" was a badly defined concept;
> 
>     b) they aren't link-local;
> 
>     c) therefore "global" scope was the only option.
> 
> 
> 
> I realise it may have just been an oversight at the time, however creating something like a "local-network" scope for ULAs, greater than link-local yet smaller than global would have aligned ULAs as scope equivalent to site-locals and then preferred over GUAs.

https://www.rfc-editor.org/rfc/rfc3879#section-2.5 still applies, I think, even though it was published 19 years ago.

    Brian

> 
> RFC3484/RFC6724 use multicast scopes to provide more granular unicast scopes, so in that case the "Organization-Local" scope would probably be best for ULAs.
> 
> 
> 
> 
>     In practice all "global" means is "not link-local" or alternatively "routeable" or "forwardable".
> 
>     The mess was somewhat clarified by RFC 8190 (part of BCP 153) defining "globally reachable": https://www.rfc-editor.org/rfc/rfc8190.html#section-2.1 <https://www.rfc-editor.org/rfc/rfc8190.html#section-2.1>
> 
> 
> I remember that discussion.
> 
> One realisation I had was that there is also a "globally unique" (or likely globally unique) property of a prefix or an address.
> 
> Site-locals weren't a globally unique or globally reachable prefix. Not being globally unique is where the issues with site-locals come up, which is also where we have issues with RFC1918s. (I.e. RFC 1627 isn't really about global reachability or routeability, but rather the lack of global uniqueness of network 10.)
> 
> ULAs have (likely) global uniqueness even though they don't have global reachability.
> 
> Link-Local Addresses don't have global uniqueness or reachability, and of course their reachability is limited to a link.
> 
> The scope IDs/interface IDs that need to be used with LLAs to distinguish them/disambiguate them is really the equivalent of the site-local site IDs that were attempted to be used to resolve the lack of uniqueness problem that site-local had.
> 
> "Unique Link-Local Addresses", with a dynamically generated and managed random portion between /10 and /64 could avoid the need for a scope ID/interface ID with a LLA prefix to disambiguate it. Just needs to be a way to have the first device on a link generate and announce it in a PIO via an RA with a Router Lifetime of value of zero, and a mechanism to have any of the other hosts take over announcing that Unique Link-Local prefix if the current announcing one disappears (somewhat like the OSPF Designated and Backup Designated router mechanism).
> 
> (Apple look to have done something like that with the Apple TV and a ULA Prefix. I don't have a packet capture, and it isn't stated, however I'm guessing that the RAs being emitted by the Apple TV have a zero router lifetime so it isn't considered a default router.
> 
> New Apple TV 4K acting as router and giving out IPv6 (ULA) addresses
> https://discussions.apple.com/thread/252823422 <https://discussions.apple.com/thread/252823422>
> 
> )
> 
> Regards,
> Mark.
> 
> 
> 
> 
> 
>           Brian
>