Re: AW: Proposal for Remote Procedure Call extensions to VISS.

On Fri, 2020-02-07 at 16:24 +0000, Schildt Sebastian (CR/AEX1) wrote:
> Hi Gunnar,
> 
> To be clear, my earlier enthusiastic message was about the need for a
> robust RPC mechanism

Understood, and I would imagine there are several.  WAMP is one that
interoperates well with web technologies due to JSON messaging being
its core only from  there considering other optimizations.
(And only staying with JSON encoding, there is also an official JSON-
RPC, as yet another example...)

> , not about a specific way to implement it. And I stand by it :)
> 
> Being a little late to this party, I have never heard about WAMP (an
> acronym which unfortunately a bit overloaded....), 

Agreed, you need to use the sites I linked to or add "protocol" to any
web search in order to avoid the Windows-Apache-MySQL-Perl stack.

> Do I understand it correctly: Your idea would be use WAMP as
> "protocol"/technical scaffolding, at least for the websocket part 

Sure.  Or parts of it, at least minimizing differences and creating the
potential for growth later on.  For starters, I think the pub/sub
message format is likely to be semantically more or less equivalent
with the VISS spec.

Then, this includes the idea of building on something that has the
potential to grow to additional features when needed.  And that all of
those features are already built in.   So you need one technology and
spec rather than piecing together multiple technologies.

> and just figure a way how to put the VSS data model behind it

I imagine figuring that out would be not much more than what VISS
already did, only with the message format that WAMP defines.  One JSON
format slightly tweaked for another, so that it fulfils the spec.

>  (e.g. defining some standardized CALLs such as viss.get / set that
> should be implemented on top of WAMP)?

There might be some semantic difference but I think you won't need to
consider it "on top of".  I would imagine it to *be* WAMP messaging, or
some subset of it.  WAMP has the defined protocol for get, set
(publish) and subscribe over WebSocket (or other).

Some semantic differences might occur since WAMP is more powerful than
what VISS spec currently requires, not only because of the existence of
RPC which is a side note, but also for pub/sub WAMP has abstract
addressing and many-to-many relationships, which are resolved by a
router***.

But on the other hand I expect if VISS/"Gen2" wants to keep the direct
peer-to-peer model, that would reduce down to a simplified version of
the same.  We won't know if we don't look.

***This router/broker model is of course completely analogous to other
popular protocols, like MQTT!   OK, I will stop there.  Actually I wasn't planning to explain or promote particular details now, only highlight that this needs to be looked at!  (Again!)  From there we can figure out the details together.

My primary message here is that the automotive industry keeps missing
the point of reusing rather than reinventing.

When it comes to learning everything about similar protocols, GENIVI
ran, for most of last year and the year before, a seriously ambitious
analysis project to look at and compare all types of similar transport
protocols and (this is the ambitious part) strive to reduce
fragmentation in the whole industry.  We have told people about this,
it has been presented in member newsletters, on publicly available
mailing lists, and so on.  A good handful of companies did the work,
but there certainly could be more.

Companies (or just parts of companies, due to that the larger companies
have trouble being fully aligned/informed across departments) have a
challenge but they are also not doing enough to fix this.

In my view, those that miss participation in such projects become "late
to the party".

The goal of GPRO project to reduce fragmentation and agree on, if
not *one*, at least a *handful* of prioritized technologies was only
partly met, in my opinion.  But we have continued to press forward with
those that were identified as preferred in the surveys that were held. 
The automotive industry is still large.  And fragmented.

> No opinion expressed here (I am missing data :) ),  just want to make
> sure, I understand it correctly.
> 

I get it.  And yes, we all lack data and knowledge, and the way to
overcome that is to participate in the collaboration opportunities that
are provided, and to really seek consensus on fewer chosen
technologies!

All the best,
- Gunnar

-- 
Gunnar Andersson <gandersson@genivi.org>
Development Lead
GENIVI Alliance

Received on Friday, 7 February 2020 17:23:56 UTC