On Thursday, July 21, 2011 9:36 AM Antoin Verschuren wrote:
> On 21-07-11 14:15, Stephan Lagerholm wrote:
> >
> > I suggested to use another flag for the DNSKEY a few months ago here.
> The
> > reaction from the list was that it was too complicated because the
> key-id
> > would change during the lifetime of the DNSKEY.
> 
> Why would you look at the key-id ?
I wouldn't, but the feedback was that we shouldn't mess with the flag field
of the DNSKEY, RFC5011 did that and it created a lot of confusion for
operators because the key-id changes.

> If as a parent I get a notify, I (securely) look up the DNSKEY set,
> select the keys with a SEP bit and hash them to DS. What does the key-
> id do in that ?
Nothing, but the SEP = DS doesn't work. Some zones (.NU) doesn't even have
the SEP bit.

> If the key-id changes during the DNSKEY lifetime, doesn't it mean the
> DS at the parent should change too ?
> Or do you mean there's a point in time you set the flag, and that
> causes the key-id to change ? Yes, sure. But then I see that as a key
change,
> and you have to signal the parent anyway.
The idea was to have an embryonic flag. For example, if the embryonic flag
was in position 10 in the flag field (512):
The child operator would publish a key with flag 769 (257 + 512) signaling
the intention to soon add a DNSKEY with flag field 257.
The parent would pick up the DNSKEY and create a DS but without including
the embryonic flag in the DS calculation.
The after the child sees the DS at the parent, it would remove the embryonic
flag from the DNSKEY.

During the lifetime of the key it would have different flag fields and
different key-ids.
769 -> 257 -> 385 (if RFC5011 rollovers are used, the revoked bit could
perhaps also be extended mean that the DS should be removed from the parent)

/S

Attachment: smime.p7s
Description: S/MIME cryptographic signature

_______________________________________________
DNSOP mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/dnsop

Reply via email to