Re: [core] draft-hartke-t2trg-coral-reef

[Klaus Hartke <[e-mail]>]

Hi Jim!

> Looking at doing more implementation work on this

❤️️:heart:

> 1.  Looking at section 5.2.5 - "Create Resource Registration"  I am going to
> assume that you are keeping the same command line as what the current
> document does, but I wonder if a better way to do it would be to use the
> following as a resource
>
> #using <http:/coreapps.org/reef#>
> rd-unit <coap://[2001:db8:4::1]/> [
> ep "node1"
> domain "Floor 1"
> lt 3600
> et "LightBar"
> rd-item <light/left> [rt "light" ct 0]
> rd-item <light/middle> [rt "light" ct 0]
> rd-item<light/middle> [rt "light" ct 0]
> ]
>
> This has the advantage that all of the query parameters are moved into the
> content which is going to make things easier to have new unknown items show
> up.  I would think that it is just fine to restrict this to a single
> rd-unit, but for third party registration it might be reasonable to allow
> multiple rd-units to occur in a single body.
>
> 2.  A registration update could then have either the same as the content in
> 1 - or just an array of rd-items.  In the first case the attributes
> associated with the endpoint would replaced - except for the ep and domain
> attributes, in the second case just the items associated with it would be
> replaced.

Moving query parameters into the content to make it easier to have new
unknown items seems like a good idea.

I'd have a more general question though: What are really trying to
model in our resource directories? Is it really that interesting to
model CoAP servers and the CoAP resources hosted by them? Or wouldn't
it be much more useful to model light bulbs, light switches and living
rooms, or coffee machines, coffee orders and coffee cups -- how they
relate to each other and how they can be interacted with using CoAP?

(See <https://mailarchive.ietf.org/arch/msg/t2trg/Q186viXtj0-r9vWEge31yuUm35c>
for related thoughts.)

> 3.  If the CoRAL format is used, are the paging attributes still going to be
> supported to keep the document size down?   Does not say anywhere.

If I remember correctly, there was a discussion on this a while ago
that came to the conclusion that block-wise transfers might be good
enough here. If that's not the case, let's have another discussion.

Klaus