3 ms·

I want it to be simple and well tested so it doesn’t need updates, just like the physical parts of the car.

by _yb2s 4y ago

fowtowmowcow 4y ago
Even with the best programmers and safety measures in the world NASA still had to update the space shuttle. https://www.nasa.gov/mission_pages/shuttle/flyout/flyfeature_shuttlecomputers.html https://www.nasa.gov/mission_pages/shuttle/flyout/flyfeature...
- 2rsf 4y ago
  And the Curiosity Mars Rover
- jimnotgym 4y ago
  That's OK, they can plug in and update during the annual service.
- _yb2s 4y ago
  I expect a car manufacturer to do extensive real world testing in all possible conditions- heat, traffic, etc. before selling it. A spaceship needs to work perfectly the first time, the actual use is also the only real world test of the prototype. It’s also a much harder problem… cars are much simpler and well understood “mature” tech where we have over a century of real world experience on what type of failures are possible and how to mitigate them.
  - fowtowmowcow 4y ago
    You know what's also cool. A mechanical engineer can lose their license if their work results in deaths. A software engineer can't lose their license because they don't have one. Perhaps a professional license may improve software development and thus better software in consumer products like cars where bad software design can result in deaths.
    - _yb2s 4y ago
      I don't think it would need to apply to all SWEs, but would make sense for safety critical things like firmware that runs vehicles, roads, bridges, etc. Certainly the sort of systems that interact with or control other systems already requiring a PE. BioE doesn't have anything like that yet either.
      - fowtowmowcow 4y ago
        I agree. https://en.wikipedia.org/wiki/Biomedical_engineering#Training_and_certification https://en.wikipedia.org/wiki/Biomedical_engineering#Trainin... Looks like while there is not a particular license for BioE they can be licensed under other disciplines.
2rsf 4y ago
Simple means no advanced safety, comfort and other features, once software is involved it will never be tested enough especially at the price point cars are sold at.
- _yb2s 4y ago
  Exactly. I have zero interest in "advanced features" which is why I still drive vintage high end european cars I maintain myself, and not a Tesla. Nearly all cars since the mid 70s have had plenty of very simple software on board in each of the control modules- ABS, fuel injection, ignition, etc. and it works flawlessly in widespread global use for decades with no issues because it is so simple. These things are so simple it's virtually impossible to have a flaw that can't be found in extensive on road testing. In very rare cases where a flaw is found, it is updated at dealerships via recall. Easy quick and frequent updates done modern silicon valley style are dangerous and absurd for something as safety critical as a vehicle. Things are practically guaranteed to intermittently break while you own the vehicle, because there is no way the extensive factory testing is repeated for each minor update.
  - 2rsf 4y ago
    Have you read the NASA investigation of the Toyota’s unintended acceleration accidents? It’s millions of lines of not properly tested spaghetti code. But I admit that it mostly worked pretty well. Unfortunately modern cars are significantly safer than vintage ones so we are stuck with complex software for better and for worse
    - _yb2s 4y ago
      I have not, but if I remember correctly, it ended up being a physical problem (floor mats) and not a software problem? Most vehicle ECUs are traditionally made with very simple C or assembly software that simply maps inputs to outputs via a lookup table... and the same basic ECU firmware is used for decades and shared by numerous car manufacturers (e.g. almost all cars use BOSCH fuel injection ECUs). If they have a complex algorithm or model to generate these tables, that code never goes on the ECU itself, to minimize the complexity and make it reliable and predictable. So the car manufacturer doesn't even develop any code, they only produce a table of numbers for flashing into the ECU, eliminating the possibility of algorithm errors. Once you physically test in diverse conditions that hit every possible value in the lookup table, there is very little room for an unexpected problem to pop up later on. Recalls updating the ECU tables have been mostly to reduce emissions and improve performance, but the vehicle still works fine under all conditions without it. One thing I find really fascinating, is that old fashioned diesel engines had a mechanical injector pump, which was basically a sort of simple mechanical computer that in some sense works the same as described above. All vehicles pretty much used the same pump, but each had different "lookup tables" physically machined into a steel block as a curved ramp. I like the philosophy explained in this article, that mentions Dieter Rams thoughts on the Porsche 911 design: https://www.classicdriver.com/en/article/design-furniture/minimalist-functionalist-influential-designs-dieter-rams https://www.classicdriver.com/en/article/design-furniture/mi... I don't buy the idea that this philosophy can't be applied to make safe and modern electric vehicles.

I want it to be simple and well tested so it doesn’t need updates, just like the physical parts of the car.

by _yb2s 4y ago

I want it to be simple and well tested so it doesn’t need updates, just like the physical parts of the car.

Even with the best programmers and safety measures in the world NASA still had to update the space shuttle. https://www.nasa.gov/mission_pages/shuttle/flyout/flyfeature_shuttlecomputers.html https://www.nasa.gov/mission_pages/shuttle/flyout/flyfeature...

2rsf 4y ago
And the Curiosity Mars Rover
jimnotgym 4y ago
That's OK, they can plug in and update during the annual service.
_yb2s 4y ago
I expect a car manufacturer to do extensive real world testing in all possible conditions- heat, traffic, etc. before selling it. A spaceship needs to work perfectly the first time, the actual use is also the only real world test of the prototype. It’s also a much harder problem… cars are much simpler and well understood “mature” tech where we have over a century of real world experience on what type of failures are possible and how to mitigate them.
- fowtowmowcow 4y ago
  You know what's also cool. A mechanical engineer can lose their license if their work results in deaths. A software engineer can't lose their license because they don't have one. Perhaps a professional license may improve software development and thus better software in consumer products like cars where bad software design can result in deaths.
  - _yb2s 4y ago
    I don't think it would need to apply to all SWEs, but would make sense for safety critical things like firmware that runs vehicles, roads, bridges, etc. Certainly the sort of systems that interact with or control other systems already requiring a PE. BioE doesn't have anything like that yet either.
    - fowtowmowcow 4y ago
      I agree. https://en.wikipedia.org/wiki/Biomedical_engineering#Training_and_certification https://en.wikipedia.org/wiki/Biomedical_engineering#Trainin... Looks like while there is not a particular license for BioE they can be licensed under other disciplines.

2rsf 4y ago

Simple means no advanced safety, comfort and other features, once software is involved it will never be tested enough especially at the price point cars are sold at.

_yb2s 4y ago
Exactly. I have zero interest in "advanced features" which is why I still drive vintage high end european cars I maintain myself, and not a Tesla. Nearly all cars since the mid 70s have had plenty of very simple software on board in each of the control modules- ABS, fuel injection, ignition, etc. and it works flawlessly in widespread global use for decades with no issues because it is so simple. These things are so simple it's virtually impossible to have a flaw that can't be found in extensive on road testing. In very rare cases where a flaw is found, it is updated at dealerships via recall. Easy quick and frequent updates done modern silicon valley style are dangerous and absurd for something as safety critical as a vehicle. Things are practically guaranteed to intermittently break while you own the vehicle, because there is no way the extensive factory testing is repeated for each minor update.
- 2rsf 4y ago
  Have you read the NASA investigation of the Toyota’s unintended acceleration accidents? It’s millions of lines of not properly tested spaghetti code. But I admit that it mostly worked pretty well. Unfortunately modern cars are significantly safer than vintage ones so we are stuck with complex software for better and for worse
  - _yb2s 4y ago
    I have not, but if I remember correctly, it ended up being a physical problem (floor mats) and not a software problem? Most vehicle ECUs are traditionally made with very simple C or assembly software that simply maps inputs to outputs via a lookup table... and the same basic ECU firmware is used for decades and shared by numerous car manufacturers (e.g. almost all cars use BOSCH fuel injection ECUs). If they have a complex algorithm or model to generate these tables, that code never goes on the ECU itself, to minimize the complexity and make it reliable and predictable. So the car manufacturer doesn't even develop any code, they only produce a table of numbers for flashing into the ECU, eliminating the possibility of algorithm errors. Once you physically test in diverse conditions that hit every possible value in the lookup table, there is very little room for an unexpected problem to pop up later on. Recalls updating the ECU tables have been mostly to reduce emissions and improve performance, but the vehicle still works fine under all conditions without it. One thing I find really fascinating, is that old fashioned diesel engines had a mechanical injector pump, which was basically a sort of simple mechanical computer that in some sense works the same as described above. All vehicles pretty much used the same pump, but each had different "lookup tables" physically machined into a steel block as a curved ramp. I like the philosophy explained in this article, that mentions Dieter Rams thoughts on the Porsche 911 design: https://www.classicdriver.com/en/article/design-furniture/minimalist-functionalist-influential-designs-dieter-rams https://www.classicdriver.com/en/article/design-furniture/mi... I don't buy the idea that this philosophy can't be applied to make safe and modern electric vehicles.