Re: [spring] IPv6 Addresses and SIDs

[Ron Bonica <[e-mail]>]

Robert,

Yeah, there were a few typos in my original message. What I meant to say was:


  *   If a /64 contains a SID, it MUST NOT contain any addresses that represent interfaces.
  *   If a /64 contains an address that represents an interface, it MUST NOT contain SIDs.

If we don't do this, we have to specify how nodes behave when they receive ICMPv6 NS messages in which the target is:


  *   A locally instantiated SID
  *   A SID learned from the IGP

                                                                      Ron


From: Robert Raszuk <[e-mail]>
Sent: Sunday, October 13, 2019 6:57 PM
To: Ron Bonica <[e-mail]>
Cc: SPRING WG List <[e-mail]>
Subject: Re: IPv6 Addresses and SIDs

Hi Ron,

/64 prefix is a pile of addresses ... if someone would be to follow your suggestion I could not allocate some blocks of that prefix on R1, then some other blocks on R2 then yet more on my servers.

You said:

"With a /64, if one /128 represents an IPv6 interface, as described in RFC 4291, all /128 MUST either:


  *   Represent an IPv6 interface, as described in RFC 4291, or
  *   Be unassigned"

Maybe you meant to say something else:

"When a /64 is used as SRv6 locator prefix, if one /128 represents an IPv6 interface, as described in RFC 4291, all /128 MUST either:


  *   Represent an IPv6 interface, as described in RFC 4291, or
  *   Be unassigned"
But then you sent this to SPRINT indicating that 6MAN should be the audience :).

Best,
R.


On Mon, Oct 14, 2019 at 12:45 AM Ron Bonica <[e-mail]<mailto:[e-mail]>> wrote:
Robert,

I'm having a hard time understanding exactly how I have violated the longest match principle. Could you provide:


  *   A pointer to a statement of the longest match principle
  *   A few words regarding how I have violated it

                                                              Ron


From: Robert Raszuk <[e-mail]<mailto:[e-mail]>>
Sent: Sunday, October 13, 2019 5:24 PM
To: Ron Bonica <[e-mail]<mailto:[e-mail]>>
Cc: SPRING WG List <[e-mail]<mailto:[e-mail]>>
Subject: IPv6 Addresses and SIDs

Hi Ron,

I disagree.

Your suggestion violates longest prefix match principle in routing.

It is huge waist of address space and is not specific to IPv6 at all.

Let me describe the deployment case where your suggestion would cause it to break:

I have /64 prefix where a few  /128s from that space I allocate to local interfaces making it a local v6 destinations on those nodes.

However in the spirit of CIDR I still want to to use some blocks of that space - say  /126 or /124 as blocks which I only use to trigger local NAT as per rfc6296. And NAT does not require local address to be a destination address so it would be a big disservice to kill such deployment option.

Many thx,
R.


On Sun, Oct 13, 2019 at 10:59 PM Ron Bonica <rbonica=[e-mail]<mailto:[e-mail]>> wrote:
Folks,

I think that we need a global rule that says:

"With a /64, if one /128 represents an IPv6 interface, as described in RFC 4291, all /128 MUST either:


  *   Represent an IPv6 interface, as described in RFC 4291, or
  *   Be unassigned"

The 6man WG will need to make such a statement since it owns RFC 4291.

                                                             Ron


Juniper Business Use Only


Juniper Business Use Only