See also: IRC log
Paul: I believe we were planning
on focusing on access and the data
... there was interest in notifications but that hasn't been
released as part of viwi yet
PatrickB: they are not yet
released but think we can in another week or so. in the
meantime we can look at media
... very different from vehicle signals
https://www.w3.org/Submission/2016/SUBM-viwi-service-media-20161213/
<paul> https://www.w3.org/Submission/2016/01/
https://www.w3.org/community/autowebplatform/wiki/ViWi
PatrickB: we shared these
descriptions in Burlingame based on this json format file
... here we are talking about the media library service
https://www.w3.org/Submission/2016/SUBM-viwi-service-medialibrary-20161213/
scribe: we always have a service
(media library), resource and here you can see artist, song,
album cover
... under medialibrary/tracks i can find all my local music
indexed
... you could do a GET on all tracks
... a track has a name, id, uri, it can have an image, links to
other items such as folders, artist, album
[demonstration on webex shared screen]
PatrickB: as you can see this is
all condusive to REST and HTTP
... it is intuitive to web developers
Paul: how did you come up with this structure, where there any influences?
PatrickB: based on common web
apis
... pretty much most music services use a similar model
Paul: I could see slight variations if people are allowed to bring their model to this architecture could be problematic
PatrickB: all the parameters
mentioned are optional, sometimes you might not have album
information
... we have this implemented at present
... you have a resource level end point that you can GET at the
end
... POST can be used when relevant
... it is possible to use to adjust ID3 information on a
track
... you can subscribe to a resource query and get a socket
back
Fulup: I was comparing with
Spotify's and it is pretty similar
... in your context how to intend to interact with those
various vendors? they will adopt your api, you to theirs, a
translator...
<fulup-iotbzh> For info Spotify API is here
<fulup-iotbzh> https://developer.spotify.com/web-api/endpoint-reference/
PatrickB: we built adapters
... Spotify can register itself as a source and put it in the
medialibrary
... it could handle queries and bring back resulting
links
... it can easily be integrated into a media browser
library
Fulup: this means you intend to have a server on the car responding to this api or in the cloud?
PatrickB: it could be either, just need the port
PatrickLue: we can have a proxy in the vehicle as well
Fulup: presently we are using @@1 in AGL reference implementation
PatrickB: there is different
levels of involvement, some things are being handled by Tier
1s
... for some systems it is our developers, and sometimes we
hand specs to a supplier of what we want
... here we had LG implement
Fulup: I would need to propose a reference implementation within AGL
Paul: it needs to be proved to
work for them to adopt it
... the code needs to be able to run against our test
suite
... for AGL they need some [linux] distro with runnable
code
Fulup: if I cannot proove that then I cannot get it accepted at AGL
PatrickLue: can we point to production vehicles as proof?
Fulup: we need workable code as a starting off point
Ted: need for implementations doesn't come up until later for W3C's spec process
Paul: you want something that people can run on as a development environment
PatrickB: we have only opened up
a more substantive than the mock server we provided to in house
developers and third parties
... we build HMI and applications in mock server environments
and then bring together into a test vehicle
... with this model instead of multiple RPC calls you can send
all the parameters you want, track, time offset, volume....
Paul: what is the feeling of other people on the call. how do we unify signals and media or do we choose not to?
Fulup: we would need to see if we
could get this could work with our identity model
... these are very different services with different
requirements
... at a low level they can use the same mechanism
PatrickB: there are use cases for wanting to be able to subscribe here too
Urata: This media library's data model could be integrated into vss data model too.
... I noticed is, in case of vehicles you have only 2, 4 or so of same type of objects such as doors, wheels etc.
... That is very different from a large media libraries in which thousands of same class object.
... I'm not sure if it is good thing to have vss and media integrated together. But of course integrating is possible.
this library's data model
Paul: I hear need for two models and separation
Fulup: there is clearly a very
different strategy on structuring the data between VSS and
ViWi
... it would be much easier to map to ViWi in AGL at present
but both are possible
... we could decide to use a similar approach in handling
media
Paul: you presently allow for
extensions and can have different data models
... due to the varied cars
... do you expect to differences in media models?
PatrickB: you need to design your
api to be flexible as Urata was saying. all the doors have the
same attributes like windows that can be lowered (or not)
... you can access all or a particular door. this works well
with object representation
Urata: Thank you
PatrickLue: we should discuss next topic before we adjourn
Paul: Fulup brought up questions of the auth model
PatrickB: you can send it the same way you would an external service with tokens
Fulup: yes but there are more complicated with the different services (spotify, iheart...)
PatrickLue: they do vary
Paul: use cases are very different for signals and media services
Fulup: I want to know how go to the lower level signals as that approach is too varied
Paul: I am not against it
PatrickLue: it seems out of
scope. the lower level protocol is so varied by manufacturer
and why we are even trying to coordinate at a higher
level
... we can provide examples
Paul: we have some of these sorts of conversations with Genivi
PatrickB: we many different
platforms across VW Group
... CAN vary widely, some have MOST
... different from model year even
Fulup: agree it is very hard to try on a lowel level
Paul: we need two implementations for W3C process and those could be AGL and Genivi
PatrickLue: I don't see a clear path yet towards standardization
Paul: this tf is to make a recommendation to the WG on possible path forward. we have a few technical topics worth covering such as auth
Ted: notifications and
identity
... and then start on proposal
PatrickB: we could decide to advance some of the other domains such as media while VISS progresses
<fulup-iotbzh> AGL adopted OpenXC for CAN low level signal mapping. We need something to implement application high level API. As today ViWi looks the easiest path as it is very close of current AGL model. If people are interested in proposing code, I would be more than happy to work on a draft implementation.
PatrickB: including possibly an implementation
Ted: we aren't required to produce an implementation ourselves but could
Paul: if people are willing to implement then you can more easily entice others
Fulup: that is something we can collaborate on
Paul: I encourage people to put together their thoughts for a proposed path for the next call
Urata: I wanted to discuss
scheduling
... we can perhaps discuss integrating VISS and ViWi at the
Genivi F2F meeting in May
... there we can decide our strategy going forward
... (about 2Generation spec, not about current WebSocket based spec. To avoid misunderstanding :-) )
Paul: correct, that was the timeline people found acceptable