Email or username:

Password:

Forgot your password?
Top-level
Fabio Manganiello

@szescstopni @rrwo @Threadbane I've been writing code at professional level for most of my life. I can confirm that I've never used generators :)

My point was about the efficiency of code and how deeply it's tested against all possible use cases - such as "somebody at some point will tilt the antenna by a few degrees and we need a fallback mechanism to find earth again".

I don't feel like there's so much depth, throughout testing of all the possible things that can go wrong, and long-term thinking in the code we write today. The code we ship is often the result of trade-offs - like "we need to release it by this date, and it's ok to cut some corners, decrease test coverage, or have errors - as long as they're below this SLO".

And I'm talking of FAANG-level code, not of the small local startup.

At ESA and NASA maybe they still do things differently (I hope), but most of the commercial software is definitely far from the level of "you can run it for the next 50 years and more, and it'll just work".

Also, this whole masterpiece of engineering could fit into 69 KB of RAM. I can't think of a single non-trivial piece of code today that can fit into that size. It's like, as we added more resources, we just started using all of them to do the same things, rather than doing more things with more resources.

5 comments
DiscreetSecurity

@blacklight @szescstopni @rrwo @Threadbane Don't worry, there's still microcontrollers, if you want to play with tiny RAM sizes! (though *some* are Gb+ of Flash, now)

Choong Ng

@blacklight @szescstopni @rrwo @Threadbane FAANG is mostly advertising and retail, it's more fair to compare the goals of software engineering at those companies to Don Draper style martini-fueled advertising campaigns and people managing inventory in the stockrooms at Macy's. I suspect the software flying modern satellites is built to an even higher standard than what's we had in the 70s.

ggdupont

@blacklight @szescstopni @rrwo @Threadbane building software has always and is still about choosing an engineering point between fast delivery and "perfectly safe code". In satellite, there is a very specific engineering point since errors are so costly. Same as for rockets or aircraft avionics...

In most other industries, the costs balance is very different. Simply because it's cheaper to pay for a car crash vs waiting for perfect guidance system in consumer cars.

Riskable ⌨️👌🏼

@blacklight @szescstopni @rrwo @Threadbane

> Also, this whole masterpiece of engineering could fit into 69 KB of RAM. I can't think of a single non-trivial piece of code today that can fit into that size.

I'm certain that you have tons of things in your home and your vehicle doing all sorts of mission-critical, sophisticated functions with less RAM. It's the difference between *generalist software* that runs on an operating system and embedded software that runs on a microcontroller.

David J. Atkinson #🟦

@rrwo @szescstopni @Threadbane @blacklight At JPL in the 1990’s we developed a provably-correct language and compiler for on-board fault control software. [It’s been so long I forget the lead guy’s name 😔]

The invention felt like matches that would light in a rainstorm. Confidence. Reliability is/was the MAJOR contribution of JPL software to aerospace. Ok, that and autonomy, but that’s another thread.

Go Up