Hacker News Reader

Comment by RealityVoid | Hacker News Reader

> Modern cars are incredibly sophisticated devices with many different computer systems running a lot of code. They all run real time OSes, which are quite distinct from an operating system like Windows. They have to do their functions correctly 100% of the time. Now cars are getting tons and tons of sensors. RADAR, LIDAR, visual cameras, etc., all running sophisticated algorithms.

While that is true, PC's were very sophisticated back in the day also. We're at a different level of sofistication currently, but I feel like that sophistication can be abstracted away. After all, currently you plug a GPU card, an incredibly sofisticated piece of technology in your PC, and it mostly just works. Why sould it not be the same for cars? I say this as someone currently writing firmware for cars and wondering why this does not happen. Mechanical complexity is one reason, but once that goes away, what's to stop us then? Maybe just the power of habbit.

View on HN

If you screw up your PC by tinkering with it, all you're doing is hurting yourself.

If you screw up your electric car by tinkering with it, you could easily kill people.

That's FUD and a bit of a straw man. No one is debating whether or not you should be able to personally tinker with your car (electric or otherwise) because that's going to be possible regardless of compatibility of parts. You're perfectly welcome to go attempt to to swap your engine and fail horribly right now on your 20 year old gas guzzler or 20 years from now on your high tech electric car. That's not a reason to keep parts from being interchangeable across cars/manufacturers.

The biggest argument against interoperability is speed of change. Telsa, the leader in electric vehicles, makes a lot of their own components and algorithms. Standardization in fast moving areas slows down the pace of change. We are at the dawning of the electric car age and even earlier in the self-driving car age.

Standards tend to either work well with mature areas or in areas where people need to pool resources together either for money or legal reasons (see codec standards).

Automotive SW is quite mature. The kind of information you're sending to the different control ECU's in the car are quite small and simplistic. I see no reason why you could not implement dynamic and state-of-the-art algorithms on the image recognition part and push the comparably small result output on the control bus.

Standards enable alternatives to spring up, so long as they interface properly. That helps with both innovation and competition.

> If you screw up your electric car by tinkering with it, you could easily kill people.

People tinker with ICE cars all the time, to the point of slotting full-size V8s in four-pot engine compartments. Componentisation makes standardisation possible, and control logic is generally done through fuzzy feedback loops, not hardcoded values, that's why I can swap whole brake lines, discs and pads and still have ABS/ESP work properly as long as I stay within reasonable tolerances. Also, I fail to see how swapping an electric engine with another would affect an entirely separate and purposefully isolated component such as LIDAR given proper design.

I see the point of Jalopnik here, there is an long run opportunity to be had, like we went from the C64/Apple II era to the modern Intel/AMD platform where you can pop in a GPU/sound card/screen/whatevs and have it just work in an instant because we standardised stuff and are provided with relevant information (such as TDP) allowing responsible tinkerers to do safe things. Hell we standardised stuff so much that external GPUs happens to be a thing basically out of thin air.

The very architecture of electric cars makes things such as upgrading the engine or battery pack a much more approachable affair that currently is, possibly even as easy as swapping alloys or tires, pending proper regulation (which could be much easier to achieve due to the tech's inherent simplicity compared to ICE engines and a proper standardisation effort, as well as certified compliance by entities like TÜV).

So, I gather you have never worked in an auto shop then.

It's not that rare nowadays, but the idea of a 'dry' shop is not an industry standard, per se. There are better 1/4 odds (pulled out of my ass, I have no hard stats here, just 'experience') that the last guy that worked on your brakes was on some sort of drug that's a bit 'tougher' than Tylenol. Ok, not just your brakes, but the brakes of every other driver in the fleet have the same odds.

I mean, do what you want on your own time, fine. But it don't stop at the garage doors, not by a long shot. 'Crank' is called that because it was hidden in the crank shafts of the Hell's Angels, back in the day. Here's and article on trying to keep a 'dry' shop: http://www.bodyshopbusiness.com/illegal-substances-and-their... Note the date here, not 1990, but 2015.

Still, you don't hear of too many problems with the brakes.

If you, rightly, have a problem with this, then push for local politicians to require licensing for technicians. California has a voluntary program: the BAR. https://www.bar.ca.gov/Industry/Getting_Licensed.html . Most independent shops have been pushing for these to be requirements for decades, but the dealers and AAMCO type folks have stalled them forever too, not hard to figure out why.

I'm not talking about things like mucking with the brakes. I'm talking about things like what the parent comment said:

> I say this as someone currently writing firmware for cars and wondering why this does not happen.

> currently you plug a GPU card, an incredibly sofisticated piece of technology in your PC, and it mostly just works.

First of all, when it comes to self-driving technology (where GPU-like devices are being used today), "mostly" is nowhere near good enough.

Secondly, do you really expect drivers (the software kind, not the meat-sack kind) and hardware to have any better results in a car than in a PC? Imagine the future where you upgrade your car's processors, but the new drivers conflict with the drivers for your airbag sensors. So now you have to upgrade those drivers as well to wait for a path from NVidia.

Sorry, this sounds nice in theory, but I don't think it's workable at this level. Batteries and motor wiring? Maybe, but packaging restrictions would take away a lot of designers freedom to design...

> Secondly, do you really expect drivers (the software kind, not the meat-sack kind) and hardware to have any better results in a car than in a PC?

A modern car is a series of different PC's connected by a CAN BUS. It's much easyer to standardise comm protocols and CAN messages than a whole PC stack. And yet, currently, there is no such common standard. I see no reason why CAN modules could not be plug&play.

There is a common SW stack standard, AUTOSAR, but it's, most of the time not firggin working or the implementation is incompatible with other implementations.

Auto industry SW is a shithole of inneficiency.

Reply on news.ycombinator.com