Meeting minutes
previous minutes
<cris_> May-30
CA corrected some typos
Minutes approved.
aligning Discovery spec with Scripting
<kaz> WoT Discovery draft
McCool: the Discovery spec needs a section for CoAP (for self description)
… otherwise is pretty much done
Discoverer section
The spec assumes CoAP is supported, but not yet
McCool: the Discoverer section will lists requirements for introductions by clients
… 1. at least one introduction mechanism MUST be supported, e.g. a URL
Zoltan: these are the capabilities the runtime needs for discovery (i.e. Scripting implementation)
McCool: this is related to who is doing discovery
… any introduction, including mDNS or CoRE is enough
… even for direct discovery, the hostname is needed
… 2. Discovery MUST find at least one TD
Zoltan: why is this one needed?
McCool: the fetching capability is needed
… could be a file as well
Cristiano: details for files?
McCool: file support might go away because of the URL quirks
… the protocols are mentioned below: HTTP, CoAP, (file)
McCool: presents diagram for discovery
Cristiano: so we should define algorithms and tests in Scripting
Zoltan: you need a priori knowledge for URLs
McCool: one can also use mDNS etc
Zoltan: this is relevant because we might want to reintroduce multicast (mDNS like) discovery
McCool: multicast is probably a local discovery method, for security reasons
… it might be behind a well known single URI
Zoltan: if we dont' reintroduce multicast, we could say the empty arguments to discovery means the impl can use any available methods it can figure out
McCool: the client needs to make distinction between local network discovery and the rest
Cristiano: there are 2 kinds of apps: ones that know the URLs, but others don't know where to get these from, and would delegate to the runtime
Zoltan: MMc made the point that no args be limited to the local network, for security
McCool: yes, and moreover, a given failed URL could be used by the impl to figure out eventual discovery methods
Cristiano: what if the runtime is provisioned with a directory? is that still allowed?
McCool: it gets in the management API, outside of Scripting
… a fallback directory might make sense
Zoltan: that is a choice the solution has to make
McCool: yes, and the impl provides mechanisms for that
Cristiano: so in node-wot we'd support these mechanisms
McCool: yes, so when no args are given, the runtime defaults can be used
Zoltan: we need references from Scripting algorithms to Discovery spec clauses
Cristiano: yes, thinking the same
McCool: we were thinking to define also flowcharts (algorithms) in addition to the pictures
Cristiano: we can prototype the algorithms and they could be also used in the Discovery spec
McCool: if we follow Thing links, it may enter into loops
… there is the requirements to avoid loops
… since id's are optional, impl needs to attach a hash
… however, the impl can make decisions whether to follow links
Zoltan: the discovery provides TDs, and whether it contains links, it's a representation problem, in the consume process
McCool: we need to think more about handling TDs with links
McCool: we'd stop with this now, but need feedback from Scripting today
… let's see of CoAP will be removed for now
Zoltan: would be nice for OCF things
McCool: we can do without CoAP, HTTP is enough for now
Cristiano: maybe Jan could comment on this
Jan: certainly, could take a look
McCool: most of this needs to be done today
… in fact it might be easier to add CoAP, so let's try to make it, make a PR for the missing section
Cristiano: we can move on
issue 403
<cris_> https://
Cristiano: should we close this or make a small PR with explicit statement?
<cris_> https://
Zoltan: the latter, more explicit is better
Cristiano: ok
issue 407
https://
Cristiano: we might need to replace observeHandler with propertyChangeHandler
Zoltan: good idea to sweep the spec for the use of internal slots on properties
issue 408
<cris_> https://
Cristiano: IMHO we can remove this, as the app is always controlling emitting an event
Zoltan: let's discuss that offline, too
Cristiano: please also take a look at issue 409
adjourned