Re: [dnssd] Fwd: New Version Notification for draft-sctl-service-registration-02.txt

[Ted Lemon <[e-mail]>]

On Mon, Jul 16, 2018 at 8:24 AM, Tim Wattenberg <[e-mail]>
wrote:
>
> I just read through this conversation and the latest version on GitHub.
> I think you made valuable improvements, especially adding some wording on
> where the "base-domain" comes from (DHCP/RA).
>
> Going through the draft in detail:
> - 2.3.1 first mentions SIG(0), we might want to add a reference to RFC
> 2931?
>
Done (not yet committed)


> - 2.3.2 these server most likely won't support SIG(0) first-come
> first-served -- we might want to emphasize this at that point?
>
 I added text.

- 2.3.3 (similar to 2.3.2) it might be worthwhile to point out, that these
> registrations most likely won't be secured
>

I think this is unnecessary.   In order for such an update to be done, it
would have to be accompanied with whatever authentication the server being
updated requires.


> - 2.4 general question: Is SIG(0) first-come first-serve protection
> mandatory to be used by a service which registers itself? Might there be
> scenario where a service explicitly wants to give other instances the
> opportunity to overwrite its service registration?
>

I think defining a semantics for this that would not be hugely problematic
is difficult and definitely out of scope for this work.


> - 2.4.1.1/2.4.2 for completeness: should we mention the SIG RR?
>

What would we say?


> - 2.4.2 as briefly discussed yesterday: should we mention that the order
> of of the update-statements does matter? I mean it's clear if the reader is
> familiar with DNS update, but I thinks it doesn't hurt to also point it out
> in this draft.
>

I added this:
  Order matters in DNS updates.  Specifically, deletes must precede adds
for records that the deletes
  would affect; otherwise the add will have no effect.</t>


> - 2.4.2 (naming issue mentioned by Toke): If I understand correctly, it
> should say "A Registration MUST include at least one Service Discovery
> update, at least one Service Description update, and exactly one Host
> Description update."
>

Yes, fixed.