I really hope we pull off some of these software adages.
Like any principle or adage, if you are experienced enough to deeply understand the reason for it, you will know when it does or does not apply.
You wont have that experience (or use it really) if you stop at your first system!
Seriously though, I agree, it's by definition that adages reveal truth. This one just feels so pessimistic and boring.
My version of this would be "the third system is the best one". Make it once, learn the space and the limitations of your solution. Make it second time, learn what isn't useful. Make it just right on the third try!
That's not what "an exception that proves the rule" means. It means "you may walk here" implies (proves the rule) there is a rule that you can't walk elsewhere.
In any event, plenty of second systems are fine. It's ridiculously outdated advice. Any time someone replaces their Flask web app with a go or rust version, qed.
I dunno what your experience is like, but I have seen this scenario repeat itself many many times in many places: someone is rewriting something without fully understanding what the old thing did, blindly proclaiming that their rewrite won't have any of the problems of that old clunky thing, introducing new bugs and limitations in the process without adequately solving the problem the old thing was addressing.
All. The. Fucking. Time. Big companies are especially bad at this.
Seems to be a variant of the phenomenon second system syndrome is meant to describe.
That's not Second System Syndrome. That's Joel Spolsky's The Thing You Should Never Do.
Also, I don't think the advice is outdated at all. I've seen it happen multiple times. However, it describes a tendency to be on guard against. It's not some sort of law of the universe.
It's also not about rewriting in Go or Rust (you may be thinking of a completely different piece of advice about not doing ground up rewrites.) The second system effect is about being overconfident from the success of an initial system with limited scope, and being too ambitious in a second system.
In my experience, it means:
"This thing stands out to us as exceptional because it is exceptional; there is a rule which normally applies, but we tend not to notice the rule until an exception to the rule reminds us of the rule by contrast".
Wait... are you suggesting that replacing a flask web app with a go or rust version is in support of or against second system syndrome? That sounds like a textbook example to me - I've seen so many second versions get built mainly for the sake of a new technology that miss a lot of the original requirements that the first version solved just fine.
(I think what the paper refers to as "versioned data" is the "time-travel queries" stuff, not multi-version concurrency control which was added after Stonebraker's time.)
My guess is the audience here on HN is less likely to run into these features in their day-to-day lives. By my observations, the HN audience tends to be generalist developer types that will readily just hand off database specialization to the most popular ORM layer for their first class development environment. And I would agree they'll not encounter these things... and if they do, they'll largely be annoyances.
I work in "enterprise" type software and I see and do a lot of specifically database development (for PostgreSQL usually these days). If we consider PostgreSQL's type system as part of the object-relational feature set, I find I make use of that fairly frequently, for example compound types... also understanding that object-relational shouldn't be confused with "object-oriented". In my own work it's primarily the more advanced typing that shows up in both procedural code I write as well as queries in certain cases. Less so, though not completely absent is the inheritance feature; there are uses for it for sure, but a rather narrow set of cases (I mean outside of partitioning, which is the most common use case).
Anyway... I'm not so sure it's really a matter of downplayed as much as it is less generally understood as PostgreSQL has become more popular.
The righteous "this time we'll do it right" is a powerful drug, and fear of missing something can drive even the most experimented people to stupidity...
What was his first system?
However, if you look at the projects he’s worked on, he has a track record of writing database engines that end up in delivered database systems.