16:49:18 RRSAgent has joined #auto 16:49:19 logging to https://www.w3.org/2021/06/01-auto-irc 16:49:21 RRSAgent, make logs Public 16:49:22 Meeting: Automotive Working Group Teleconference 16:49:57 agenda+ Transparent auth capabilities 16:50:11 agenda+ Curve logging code tour in VISS2 16:50:30 agenda+ Filters, required and optional 18:02:30 Present+ Ulf, MagnusG, Peter, Arman, Ted 18:03:01 Present+ Carine 18:05:30 zakim, take up agendum 2 18:05:30 agendum 2 -- Curve logging code tour in VISS2 -- taken up [from ted] 18:06:06 Ulf: last week or so, I've been playing with converting curve logging from open source version from Geotab's github repo 18:06:22 [Ulf shares screen, architectural overview] 18:06:58 Ulf: there is a unified protocol on what I call the transport layer handled by the manager 18:07:15 ... we only have one manager implemented but possible to have multiple 18:07:55 ... I added curve logging to service manager. I split off into a separate file since the algorithm is a bit involved to have it easier to read 18:08:36 ... function activateIfIntervalOrCL 18:09:03 ... differentiates which is used 18:09:21 ... it first looks at the signals received and divides into 1, 2 or 3 dimensional signals 18:10:28 ... a simple signal like speed that varies over time is 1 dimensional. Latitude could also be represented as 1 dimension but logically belongs with Longitude hence 2 and perhaps want a 3rd which is accelerometer 18:11:15 ... we use max error for criteria to register new data point collection 18:12:00 ... what signals you want to be tracked in combination (dimensions) can be configured in json 18:13:05 https://github.com/Geotab/curve 18:13:40 https://github.com/MEAE-GOT/W3C_VehicleSignalInterfaceImpl 18:14:11 Ulf: I may optimize code a bit further as there is a bit of redundancy 18:15:24 ... regardless of how many signal dimensions you use, typically 1, you also have time as another dimension. minimum is a 2 dimensional graph on x, y axis 18:15:47 ... on first request, freshest value is sent 18:18:08 ... in my implementation I introduced a sleep of 800ms to not overload the sqlite server used for this prototype 18:18:44 https://github.com/MEAE-GOT/W3C_VehicleSignalInterfaceImpl/blob/master/server/service_mgr/curvelog.go 18:21:07 Ulf: I build up a ring buffer at the start. the analysis is quite quick 18:21:32 https://www.youtube.com/watch?v=VJfu64gdGGo explanation of benefits of curve logic in minimizing data collection 18:22:11 some have equated it to lossy compression, reduces number of data points needed to provide a reasonably accurate depiction 18:23:09 Ulf: when buffer becomes full or an unsubscription, we break out of infinite loop 18:23:28 ... that brings us to the core logic, rest is just information flow 18:23:50 ... core logic figures out what points are needed to save and tosses out others 18:23:59 Present+ Erik 18:24:48 ... here we use a linear buffer, translate strings to appropriate variable types and use Unix timestamps 18:25:42 ... the actual reduction is performed on this buffer and max error. the method is recursive and calls itself 18:26:05 ... if you look at the Geotab python in github you'll see that represented here 18:27:14 ... line is 'drawn' and compared to max error for the various data points to see if any sample exceeds max error 18:28:53 ... what this method returns is an array of the indexes that should be saved 18:30:46 ... rest is basically correcting sorting since the recursion interferes with that. the data points corresponding to the indexes array are then collected from the ring buffer 18:34:22 Ted encourages people to comment on met topic of optional versus required filters https://github.com/w3c/automotive/issues/381 18:34:28 zakim, take up agendum 1 18:34:28 agendum 1 -- Transparent auth capabilities -- taken up [from ted] 18:38:11 Ted summarizes discussion from last week and captured in comment on https://github.com/w3c/automotive/issues/382 18:38:42 Ulf: we would like to be able to accomodate what Bosch is asking for, a less complex authentication scenario 18:39:24 ... my idea is we do away with the access grant server in the cloud, claim and receive a role 18:40:02 ... then an access token is received in the vehicle. my understanding is Bosch only has one server in the vehicle 18:40:44 ... this can be handled by access grant server issuing a token with no expiration date. this can be provisioned in any number of ways without even a server 18:41:27 ... this simpler scenario removes access grant server entirely and essentially matches Bosch's model 18:42:10 ... this is a claim and although Isaac, Ted and I believe it is feasible, need to prove 18:44:00 ... the other major difference is role based authorization and there think we have a superuser role with granular exceptions 18:44:05 Present+ Gunnar 18:45:31 [Gunnar leaves] 18:46:29 Erik: Sebastian has better knowledge of what we have implemented 18:46:58 ... access grant and token servers in the spec are themselves not normative, right? 18:47:11 Ulf: correct, it is currently not mandatory 18:47:41 ... currently it is all or nothing 18:50:22 Erik: would access token be required to be short lived? 18:50:40 Ulf: you have a token and it must correlate to some extent? 18:50:59 Erik: intention then is access token should still be there 18:51:30 Ulf: yes and assume you generate tokens somewhere/somehow in your system 18:52:39 ... access token could be integrated in VISS implementation or kept separate. I kept them as separate servers in my implementation 18:52:56 Erik: I will check with Sebastian and others, get their opinion 18:53:52 Ulf: we are willing to make tweaks but idea that I hope we can achieve is we keep as much of the architecture and handle both scenarios 18:58:27 Ted: first and foremost we want the client/VISS interactions to be consistent, architectures may vary some and either or both solutions may need some tweaks to meet in the middle to attain that 18:58:42 ... we should do that first before getting into roles 18:59:00 Erik: I'll discuss on our end 19:00:10 rrsagent, draft minutes 19:00:10 I have made the request to generate https://www.w3.org/2021/06/01-auto-minutes.html ted