Re: [dnssd] SRP Update - removing individual services (draft-ietf-dnssd-srp-06)

[Abtin Keshavarzian <[e-mail]>]

Thanks Ted,

Yes, that makes sense and sounds simpler. :)
I think it also allows multiple adds and removes, all, to happen from the
same SRP update message.

Let me try to describe some details:

- In "Service Description Instruction", we allow a "Delete all RRSet from a
name", with possibly no follow-up adds.
- When a service is being deleted, we can consider allowing KEY RR "Add to
an RRSet" for instance name to be optionally included, to have the server
keep the instance name reserved (if desired).  Does it make sense?
- In "Service Discovery Instruction" I think we need to allow delete as
well (I guess "Delete an RR from an RR set" for PTR RR)?
- We continue to require SRP Update msg to always include exactly one "Host
Description".

Does this all sound good/reasonable? Anything I may have missed?

Sorry I think I misunderstood the earlier discussion. Somehow I associated
the remove to require registering with zero lease time.

Thanks,
Abtin.

On Thu, Dec 17, 2020 at 3:35 PM Ted Lemon <[e-mail]> wrote:

> Thanks, Abtin!
>
> To be quite frank, using a zero lease time was a quick hack, which only
> does well if the only use case is that the whole device is going offline
> and hence wants to take all its services offline. This is a fairly common
> use case in mDNS, but I hadn’t really thought about the fact that your use
> case also makes a lot of sense, and isn’t properly addressed by my quick
> hack.
>
> For the use case you are proposing, I think the approach we agreed on in
> the off-list conversation you are referring to indeed makes more sense. One
> slight clarification, from your second point in your summary: the lease
> time would be the normal lease time.
>
> What would be different would be that a delete would appear in the message
> with no following add. This is sufficient to delete the old service. It
> would be permitted under the conditions you describe. The new service that
> was added, and the host record being renewed, would have the new lease
> time; the old service would be immediately removed.
>
> Does that make sense??
>
> On Dec 17, 2020, at 5:55 PM, Abtin Keshavarzian <[e-mail]> wrote:
>
> Hi all,
>
> Would like to summarize some questions/discussions we had on an email
> exchange (with Ted, Kangping, Jonathan, and Rongli) related to SRP and the
> process for removing individual service(s).
>
> First a quick (personal) note, I have been reading the RSP spec recently.
> I think it is a very well-written and easy-to-read/follow. So I want to
> give thanks and kudos to guys who were involved (Ted, Stuart, others).
>
> -------
>
> There may be use-cases where we want to remove a previously
> added/registered service. The question is how to realize this.
>
> I think this can be done by sending two SRP Update message:
> - First one removing all services/host-info (with lease time zero)
> - Followed by another SR Update re-adding services (excluding the ones
> removed) and host-info
>
> However, it'd be good if this can be done more efficiently with a single
> SRP update message.
>
> - Currently SRP considers the message to be valid if it includes zero or
> more Service Discovery and Descriptions, and exactly one Host Description.
> - The idea is to extend spec to allow Update msg to include Service
> Discovery/Description (without Host Description) with lease time zero to
> use for removing services.
> - The update message needs to include the key RR and be signed.
> - The server would accept the removal of service only if the key matches
> the previously registered key associated with the host and the service, and
> if the signature is valid.
>
> Thoughts?
>
> Abtin.
>
>
>