Re: [v6ops] draft-ietf-v6ops-ula-usage-recommendations - work or abandon?

[Mark Andrews <[e-mail]>]

In message <[e-mail]>
, Lorenzo Colitti writes:
>
> My recollection is that we never got around to publishing this document
> because we never managed to agree on what to say in the recommendations. I
> think that because the working group does not actually have consensus on
> what ULAs should be used for.
>
> Generalizing and handwaving a bit, I think that the major disagreement is
> between:
>
>    1. Some proponents of ULAs are suggesting that ULAs (are being | can be
>    | should be) used in similar ways to RFC 1918, including using them behind
>    NPT66 or in the absence of global addresses, and say that the document
>    should list those as use cases.
>    2. A fair number of WG members are vehemently opposed to such uses, and
>    say that WG documents should explicitly call out such practices as
>    harmful. We heard a few of those members at the mike yesterday during the
>    discussion of the design choices draft.

A lot of this is to do with how addresses are provisioned.  If they
are provising manually (PA or PI) then you can get away without ULA
as your internal addressing is not going to evaperate on you when
you become disconnected.

If your addresses are provisioned automatically then ULA is *critical*
to providing internal stability when the PA prefix goes away.  No,
we should not be encouraging ignoring lease times.

We should also demonstrate that ULA in conjuction with other GUA
is not harmful.  One way to do that would be generate a ULA prefix
in BA and only provide DNS using addresses from the the ULA prefix.
Even better would be to get a second GUA prefix and migrate the
network from one to the other over the course of the week.  Bring
up the second GUA on Tuesday and withdraw the first prefix on
Thursday.

Mark

-- 
Mark Andrews, ISC
1 Seymour St., Dundas Valley, NSW 2117, Australia
PHONE: [phone]                 INTERNET: [e-mail]