I think you’re over simplifying auto repair work. Sure, sometimes it’s an oil change, which is short, well scoped, etc... But a lot of times it’s “Sometimes, when I’m turning left, the car makes this noise... you know, like a whining grinding popping noise. It’s not doing it today though.” And you don’t know if it’s a 5 minute fix or a 5 day investigation and replacement of major drivetrain parts...

Auto repair is 90% diagnostics and 10% wrench turning. A lot of newer vehicles have symptom matrixes with a process for repair. Lowering the skill floor for mechanics.

What’s interesting to me about auto repair is how techs are paid. A lot of shops do book rate. Example let’s say you have to replace a front half shaft. Book rate pays you for 30min. If you take an hour to do it - you lose money.

HVAC repair starts getting into black magic territory. Giant, incredibly dynamic systems with hard physics, positive/negative feedback loops, control theory operating with very few inputs and very uncomfortable humans dripping sweat on the back of your neck while they watch you work.

I feel like everyone that frequents HN would have a good time spending a year working in that industry.

I've had a tangential interest in the physics / mechanics of air-conditioning since working in an office with two people on opposite ends of the "comfortable temperature" scale, and noticing that one of them opens their window in order to "solve the problem". I've since noticed various other people open windows to try to "fix" the air-conditioning temperature.

My assumption is that opening the window fucks it up for the whole system, but very few people are aware of this, and the office folks aren't made aware of it by the air-con techs.

Am I way off? Am I overstating the effect that an open window has on the system?

It will upset the building balance at a minimum. Air balancing techs are typically a different discipline from HVAC in large commercial spaces.

Leaving the window open and AC on in high humidity areas can freeze up the coil generating a service call, trick zone sensors, etc. If there is a large temperature difference the sharp gradient will not be comfortable.

I never worked on very large systems, mostly small commercial and residential. A buddy of mine supervises HVAC for a large commercial facility and we were talking about their makeup air. Basically it sounded like the building is effectively hermetically sealed until the makeup air vents open because of interior pressure or O2/CO2 levels. Seemed kind of creepy that there was some active system in charge of that. How do you know when it's not working?

The same CO2 sensors that call for make up air are hopefully monitored. Sick building syndrome.

Buildings are poorly sealed in my experience. In California code requires economizers on RTUs and they are almost always broken / stuck partially open. In the Inland Empire for example the actuators bind up from sand and often seize. Nobody wants to pay to fix just that, if they even know it’s busted.

I stay under 10 tons. Things are different in bigger buildings!

>Auto repair is 90% diagnostics and 10% wrench turning. A lot of newer vehicles have symptom matrixes with a process for repair. Lowering the skill floor for mechanics.

That’s unfortunate because nobody actually knows what is doing and everything is fine as long as it falls in the guidelines, if they found a problem that is not “in the books” they have no clue how to proceed and this problem not only applies to mechanic shops, i’ve seen the same situation in the medical field.

The skill floor is under the house: I have a 3-series BMW that was having odd engine hesitation but not throwing any codes. The dealership's mechanics could not fix it! They ended up throwing a ludicrous amount of parts at my car including some very expensive items such as a new transmission, new transfer case, new fuel injectors, plugs, coils, fuel pump, wiring harness. The bill, which was covered by the warranty, totaled over $10,000 and they still had not fixed the issue. They finally called in a guy from an Audi dealership and he fairly quickly diagnosed the problem as a faulty oxygen sensor. They had the car for nearly 60 days while I drove a brand new 4-series loaner car. So BMW did right by me, but holy crap I have never ever seen such lousy troubleshooting!

They'll just wait until it breaks. My car clanks sometimes when doing sharp turns at a low speed. Something in the steering or wheels. Not very loud, but disconcerting. 2 visits to the dealer - zero progress. To be fair, it's probably harmless. Or it could crap out at 80mph and kill me, I don't know. And that's BMW, where mechanics are expensive AF.

Is it a 'tuk tut tut tut tuk' sound? If so, try going to a parking lot and holding the wheel all the way to the left and accelerating slowly in a circle. same for the right. If it makes that noise it's most likely your CV axles. If it's only when you hit bumps then it may be a ball joint that is coming loose.

Take it to a non-dealership? My (slight) expérience with the BMW dealership is that they’re mindless flow-chart followers that can’t/won’t use their brains to think.

They certainly didn’t read my notes that indicated that I already swapped around the plugs and coils but the problem stayed in the same cylinder.

Their first course of action was to replace all plugs and coils.

Ugh.

Was it warranty work?

They are going to do it the way that makes sure they get paid.

Nope, well out of warranty.

Ultimately, they missed out on a lot of work. Actual issue was a broken exhaust valve spring.

Indy got the same notes as me, immediately did a compression test and easily knew it wasn’t plugs/coils.

Software development is more like manufacturing engineering. Auto repair is probably closer to working an operations ticket, only without all of the fancy telemetry and diagnostic tools, and a lot more bloody knuckles.

The OBD port can tell you a lot. Especially if you can google it to get pointed toward the more common failure modes that lead to code XYZ.

Lots of shops do ground up builds.

I think it's also not the case that software development is so heroic, most deployment strategies draw a line between development and the version people are using, and there is quite some power in the ability to go back to a previous version.

I'm beginning to think the real answer for software development is to stop wanting a process. A team either clicks, or it doesn't. I think this happens no matter what process is attempted.

EDIT: Obviously there should be acceptance criteria for the product, especially where safety is a concern; but the day-to-day process is just going to be what it is.

I don’t understand why some areas would warrant the response of “sometimes it works and sometimes it doesn’t”. What if we left other areas examined? “Let’s not bother to research sorting algorithms further. Some people write a sort and it goes fast, and some people write one and it goes slow, and that’s just the way it is”.

If a team can just “click”, isn’t it critically important to figure out why this occurs, so we can create other teams that click? It would be an awful shame if we were not as a species able to make effective use of all of us because we preferred to leave this to chance.

I think it's just very hard to generalize from the performance of a combination of people that each have unique personalities and abilities. It's hard to even know what to measure.

For a sorting algorithm, you know what to analyze, and you know the result is generalizable.

No one has even bothered to use Fred Brooks' recommendations in nearly 50 years, despite praising his book up and down without reading it.

Eventually everything will be sealed in lego like blocks. We will then stop calling software technology.

That’s called hardware. The whole point of software is that it’s soft.

How far away from that do you think we are and why?

It wont happen overnight (if at all). For hardware we are already there. We use a single cpu or video card that just works and cant be matched by combining a bunch of smaller components. We do have instances trying to combine the 2.

We also have lots of and continue to have more software monoliths that at first are just to good, one wouldn't bother writing their own. Later we wont even remember how to do it and end up with a final design. Could be epic, could be aweful. (I mean every ready made thing available for use either for end users or programmers)

Why? Software is a fantastic prototyping tool, its so good we don't bother making things into a chip. This for the most part because we wouldn't be able to update it. The need for updates will eventually go away. Things for now continue to be improved but at some point it stops making sense to do so. What [say] an email application should look like was completely unknown in the 80's but we all have expectations today. It seems unlikely for us to continue to write [say] new email clients from scratch forever?