<scribe> scribenick: ted
<scribe> Scribe: Ted
<PatrickLue> https://github.com/w3c/automotive/pull/288
PatrickL reviews the pull request comments
Ulf: I am not quite aligned with
the terminology yet
... there is quite a bit of discussion still on the signal node
type
... it feels a bit outdated as we have several new node
types
PatrickL: completely agree with you but want to capture what is currently represented
Ulf: ok with an intermediate step
PatrickL: we can make incremental changes as easier to digest
Ulf: there is no mention of these new extensions to the VSS data model which we are inheriting from ViWi such as the resource extension
PatrickL: I haven't given that
thought yet. it is only there for now as described in the
graphics
... an objection to me merging my pull request, it seems
sensible to ask
<scribe> … done
<PatrickLue> https://github.com/w3c/automotive/pull/289
UNKNOWN_SPEAKER: next one,
wildcard usage over http chapter added
... I confess to not having read that yet, has anyone else had
the chance?
... based on the size of the change, I would be comfortable if
you provide some background here
Ulf: I think everyone is aware of
the wildcard mechanism in VISS, allowing path description to
address several nodes
... with the current model for gen2 and wanting to ensure we
carry forward desireable features from VISS and ViWi
... for the web socket protocol it will be the same as it was
before, for http we need to encapsulate this in the request
URL
... wildcards are special/reserved characters in URL according
the RFC
<PatrickLue> Discussion from May 2018 around queries: http://www.w3.org/2018/05/29-auto-minutes.html#item01
Ulf: I have a couple ideas and we
could try to settle upon one or allow for both
... we can replace * with the corresponding ascii hex value as
that is allowed by RFC
... that makes the expression a bit unfriendly to read and
thought a bit more about it to try to come up with something
more friendly
... we could use a different character than * and chose the ~
as an alternative
... I stated wildcard usage shall be supported, required by
conformant server implementations
PatrickL: we had a discussion
around this previously, need to find the sweet spot
... what I don't like is addressing and querying combined, they
are two distinct things
Ulf: I think they can complement
each other, the query mainly to be used for resource type
nodes
... there the query mechanism is necessary
PatrickL: to me the example to
find all doors that are open, it seems more natural to get all
doors where open=true
... addressing something with wildcards is a query for me
... because I am not addressing one object but a series of
them
... I think it would be hard to express all with wildcards
Daniel: there is an open issue
with VSS regarding the positioning within the tree
... position can be encoded eg right, rear
... we were wondering about that at the end of the path
... having position at the end may facilitate
Ted: querystring may allow for
special characters that are otherwise prohibited in server/path
and other portions of a URI and would need to confirm
... we can have multiple wildcards at different points in a
tree/path which complicates things, having a partial path and then
querystring that includes different branch levels would be
complicated
... we should come up with an array of possible wildcard examples
in VISS and map them to proposed URI solutions
Ulf: I am not so focused on the mechanism but want to ensure we continue to support the feature
<PatrickLue> <service>/car/doors/?open=true
PatrickL: the ViWi style anotation would be as follows
[what i am struggling with are scenarios that would have foo/*/bar/* in VISS]
[after you go qs you can't return to address path in url]
Ulf: as long as we can get the same functionality, I am fine with any mechanism
<PatrickLue> <service>/car/doors/?row=*&side=*&open=true
PatrickL: every level has a certain idea behind it, /car/doors/?row=*&side=*&open=true for example could address Ted's point
Daniel: coming back to VSS,
positioning is awkward as I mentioned I will bring forward some
ideas
... we need to avoid breaking the tree structure. taking parts
out of the tree path and put in query would be a bit off
too
... let me work on an idea for next week
PatrickL: based in the discussion let's hold on the pull request
Ulf: absolutely, that is fine
[Agenda bashing for this and next week]
<PatrickLue> https://github.com/w3c/automotive/issues/287
PatrickL: we have not agreed yet
on addressing schema yet to my knowledge
... proposal is we want a URL style addressing converting . to
/
... Ulf is already referencing the style we know from VISS and
think we should have one addressing style throughout,
independent of protocol
<PatrickLue> https://www.w3.org/2018/08/28-auto-minutes.html#item01
Ulf: I would support that
Peter: makes alot of sense
PatrickL: people have until I
finish typing in gh to object
... next issue is... usage of wildcards...
[covered with plan to return after hearing idea Daniel is forming]
<PatrickLue> https://github.com/w3c/automotive/issues/282
Ulf: it is not followed in the current specification text, so I would like to kill the idea
agree
<PatrickLue> https://github.com/w3c/automotive/issues/281
PatrickL: the rest are non-gen2
issues
... any brief topics anyone wants to discuss in remaining
time?
Ulf: just a shameless plug to
encourage people to check out the implementation and consider
contributing to it
... servers implemented in JS, uses VISS path for search path
responding with number of matches
... it can keep a number of http and web sockets
concurrently
... it is progressing, looking for contributors
[adjourned]