Hacker News Reader

Comment by collegeburner | Hacker News Reader

This definitely doesn't match how I heard some software is done there: https://news.ycombinator.com/item?id=18463181

FWIW (i wrote that comment, also btw woa I'm super flattered to be quoted like that!): note that it's based on my, by now, pretty dated inside knowledge (~6y). They might've improved since (though judging by the sibling comments, maybe not).

I find it funny how the other commenter talked about how ASML could be disrupted by a competitor with better software practices. Completely laughable in todays chip shortage world.

I have had to learn this lesson the hard way when I assumed that Disney+ was going to flop when a lot of the senior engineering talent supposedly quit during the ramp up. What I didn't realize was that technology isn't as important as other things when you have a moat. Disney+ ended up relatively successful despite their technological hiccups. Same reality with ASML.

Definitely two types of software here. Everything about that post is spot on.

Internal software is messy. Testing is also messy. Lots of things need to be automated. End of the day, it’s a bunch of hardware people.

The exciting stuff are the physics simulations. Modeling how lens aberrations affect the light that passes through and how to correct for that.

Software wasn’t mentioned in the comment though. It might be an awesome place to work at within those other disciplines which were mentioned.

This is the sort of thing that shys me away from every working at a hardware company doing software.

yep. accurate comment. How do you like 40M LOC with 50% code duplication?

That sounds like a train wreck waiting to happen. I’m amazed that hasn’t put an egg on their face. Are their customers complaining about issues?

And how can the risk managers know so much about potential regressions? Are they a panel of former engineers?!

So many questions

keep in mind that machines like those are so complex that you don't simply buy them and install them.

It's more like getting custom-made artisanal goods: each machine has a "father/mother" from its 'conception' til delivery who treats it as their (work) baby. Then a special plane will fly the machine and a team of dedicated engineers will spend weeks in your fab doing install / validation. The process is long and complex.

Downtime are really painful and expensive so the philosophy is "better copy paste the code rather than change what's working". downtime are not very frequent but people are just the least expensive / worrying thing there. It's not like you can easily start a competitor...

It almost sounds like there should be states funding creation of a competitor simply to prevent the "fire impacts entire global industry" kind of news we're seeing today.

ASML has 16 locations around the globe so while it will have some impact, it could be a lot worse.

https://www.asml.com/en/company/about-asml/locations

Not every location does the same thing though. ASML has bought other companies that were key suppliers to them. Like Berliner Glas Group in 2020 which is located in Berlin and is where the accident happened. No other location of ASML will be able to compensate for production losses there.

a competitor won't be enough, the whole supply chain is made by companies which are single point of failure (the reason why europe can't afford to say "ASML only sell your machines to EU fabs" is that one of the key ASML supplies is US.. and without them no machines...).

Well China is trying are they not? We will probably have to wait another 5-15 years to see the results...

The customers do complain, but they only care about wafer yield and throughput. They will accept any dirty software hack or manual procedure to circumvent issues _now_.

Reply on news.ycombinator.com