<scribe> scribenick: ted
Volker: my focus is on navigation and fleet data aggregation
Glenn: we had a call from Darren
at Geotab where he described the metadata and processes for
collecting the data that data consumer will need to know
... what we thought would be useful would be the key topic
areas for such a contract
Harjot: concepts are exactly as
Glenn alluded, we want to start with an agreement on the broad
topics and will start off with listing them, please interrupt
with questions or comments
... first would be how the information is logged, how
frequently and algorithms used
... it could potentially account for error thresholds
... data scientists would preferred a curve algorithm instead
of captures at regular intervals
... who has access to what data and when which Caruso is
leading could be part of the contract as well
... if someone is using a vehicle they don't own for example
but are allowed to use it for off-hours then you may have
different models
... there will likely be other more complicated use cases as
there seems to be a shift from individual ownership to usage
based
... there will be different regional privacy legal restrictions
such as GDPR in EU
... identity is also important you want to be sure
Glenn: what we are looking for is confirmation that these are the right ones before investing resources
Benjamin: what I am more
comfortable with is attaching this metadata (auth, methodology,
etc) directly to the data itself
... I am looking forward to concrete examples
... I want to hear how it would be modeled
... beyond that I do not have further insight
Ted: have you by chance looked at ODRL yet?
Benjamin: not yet
Ted: my understanding is that it is used by media and news industry
Volker: I am still trying to get
a grip on the scope you are working on
... we looked at GDPR ourselves extensively
... trying to get data tied to a specific reason. we are also
looking at depersonalizing data
... for personal information you need to be able to get clear
acceptance from the user
Harjot: it is a fine balance between providing anonymization and aggregate data while retaining useful information
Glenn: with respect to
anonymization, it is a clear area to address
... data scientists are rather clever and can often, almost
always, reverse engineer information and be able to identify
people
... you would need multiple vehicles in the set and element of
randomization
Benjamin: there have been various
example of reverse engineering
... there have been some good practices produced
Ted: @@first_pass
Harjot: we will probably start
with a Google doc and then share with the group
... export to wiki when it gets more stable
https://www.w3.org/community/autowebplatform/wiki/Consent_Cases
https://cordis.europa.eu/project/rcn/86393_fr.html
https://lists.w3.org/Archives/Member/member-automotive/2018Jul/0007.html
Benjamin: the problem is VSS is
that it imposes a data tree structure
... I may want to have it organized in a different structure
and what I have done with VSSo
... I have generated classes and subclasses drawing from
branches. each class has a unique uri
... for signals and attributes when relevant I include unit
types
... vehicle speed in VSS, it has a label and description
... it could be in km/h or m/h
... we want to distinguish between static information and
dynamic signals
... point is to be sure we are talking about the same concepts
and people can do different mapping
... we want to make sure the resulting model is still developer
friendly
... there are only a few critical areas that need
clarification
Ted: how are they deviating from one another?
Benjamin: mostly in how they
structure the data
... VSS does a specific tree representation of a car. while it
makes sense from a certain point of view, we learned from
Caruso and ViWi they are organizing in a different way
... that stresses the need to have a flattened view and let
people structure how they need
Ted: @@viwi_sampling
Volker: we are using it but it
might not end up being the main way we will collect data
... how we represent it in warehouse will be different
... we have personalized information that can follow them
across vehicles and take a separate vehicle centric
approach
Adam: we are storing individual profiles off-board and synch to individual vehicles
https://doodle.com/poll/avrcb7thiyawtvug#table
Cancelling 9 August
[adjourned]
This is scribe.perl Revision: 1.152 of Date: 2017/02/06 11:04:15 Check for newer version at http://dev.w3.org/cvsweb/~checkout~/2002/scribe/ Guessing input format: Irssi_ISO8601_Log_Text_Format (score 1.00) Present: Benjamin Harjot Glenn Ted Volker_Remuss(Carmeq) Adam Regrets: Dominik Found ScribeNick: ted Inferring Scribes: ted WARNING: No meeting chair found! You should specify the meeting chair like this: <dbooth> Chair: dbooth Found Date: 12 Jul 2018 People with action items: WARNING: Input appears to use implicit continuation lines. You may need the "-implicitContinuations" option. WARNING: IRC log location not specified! (You can ignore this warning if you do not want the generated minutes to contain a link to the original IRC log.)[End of scribe.perl diagnostic output]