As an academic, it must be soul destroying to see the world dominated by ad-hoc efforts like Python. Just like Linux, it was one persons hobby project that repeated old mistakes and now threatens the existence of argubly better technology.
As an academic, it must be soul destroying to see the world dominated by ad-hoc efforts like Python. Just like Linux, it was one persons hobby project that repeated old mistakes and now threatens the existence of argubly better technology.
PL is like art theory & art criticism... You think more about individual aspects, and as long as something's happening, great. Ex: concurrency is super hard, and where python got to with async/await is surprisingly principled for a dyn lang. In contrast, Go's initial interfaces mess was almost an intentional snubbing that unsurprisingly had to get revisited.
Weirder is big ideas take literal decades to come out. Ex: mypy is types from 50 years ago, while stuff like pandas suggests we still need basics like dependent/row types for typing any data science code, which is not a new idea either. But with OSS, it's become more about academic hubris / ivory tower vs an indictment of pythonistas for whether those happen.
It's a pretty fun time in PL for folks who do care: easy to start free frameworks with global reach and not worry about funding/sustainability, so a lot of playing in areas like synthesis and data.
What do you mean by Python's mistakes? I'm no Python zealot but I'd say the Python language broadly achieves its aims. It's not a platform for cleverness with type theory or monads, and it's not a sensible choice for proving the correctness of programs. Those aren't its aims.
I accept that it's a huge improvement over many mainstream languages (certainly bash!), but I find e.g. OCaml/F#/ML code far easier to reason about. These languages are also hybrid imperative and so there is no need to use monads, they can be used just like Python.
As with many other fields of endeavor, it is very hard to compete against free. Throw in the ability to fix any problems that might affect you personally, and you have a powerful combination.
The same goes for Apache as well, and to a lesser extent MySQL and Perl/Python/PHP.
Trying to set up an equivalent commercial stack was both painful and expensive (and not just the software licensing, but the minimum hardware requirements too). Not to mention that once you're considering commercial options, Microsoft was a contender, often with better initial ROI (the cost to exit Microsoft platforms was quite another matter).
However Linux is free, and we have all benefited enormously from that, myself included and the tech giants.
Kernel architecture (be it micro or monolithic) per se is no virtue.
> Kernel architecture is also highly relevant, ever wonder why Android phones only get a few years of updates?
I figure in your mind it has to do something with micro-architecture rather than vendors not giving a flying fart, so I'll bite. Can you explain the mechanism by which a monolithic kernel update is any harder than a bunch of services?
No, you were wrong with the statement "This also glosses over that commercial Unices weren't superior technology.". The context was Linux killing off the (technically superior) competition, a process that took decades.
There are many other technologies that Solaris had years before Linux, e.g. decent SMP, Zones (containers), ZMF management and many other features that Linux still does not have. QNX had realtime features and a microkernel.
> I figure in your mind it has to do something with micro-architecture rather than vendors not giving a flying fart, so I'll bite.
The situation is more complex than "vendors not giving a flying fart". Linux does not have a stable module ABI, in fact, it doesn't even guarantee a stable API. This is all part of "kernel architecture" and highly relevant. Regarding microkernels, security concerns will eventually force this design on future operating systems (e.g. Google's Fuschia)
I dunno that I see it that way. (I am a PhD student in programming languages, likely going to end up in academia.)
In many ways, yes, Python is "bad". It lacks an expressive static type system, it has some bizarre rules about scope, and there are other idiosyncrasies that, were I to build my own language from scratch, I would seek to avoid.
But in many other ways, Python is very good. The standard library is phenomenally complete, featuring functionality for the vast majority of common use cases. I also find the documentation to be pretty approachable, though that sentiment seems to not be very well-shared. As for the language proper, the lack of a type system does irk me somewhat, but I use mypy to get back some of those guarantees so it's not as bad as it could be. I find the syntax mostly pleasant, and the module system is usually intuitive. Some of the things they've added are very nice features: I like that multiple consecutive string literals are treated as a single string (useful for splitting strings across lines); I like f-string formatting, too; the use of underscores as numerical separators is great (though not unique to Python); the overall design of the language, while truly odd at first, is relatively consistent and powerful (e.g., everything-is-an-object, iterators/iterables are prevalent, classes can be modified dynamically [which is maybe not always a good thing, but it's a neat tool for the power-user], etc.).
Again, it's not a language I would build myself, but as far as languages that exist, it's not even close to my least favorite.
I think it's also worth considering the space of programming language design in linguistic terms. Natural (human) languages are amazing because they constantly evolve through regular use. New terms are inducted into lexicons constantly, and new grammatical rules can be generated and regularized among a group to give the users their own characterization separate from other speakers of the same language. Natural language is beautiful.
If we look at individual programming languages as separate languages, they do not exhibit the same properties. They are pretty much static, except when a new major version is released.
But if we look at programming languages rather as dialects of a single common language (easy to do when we consider "Turing-complete languages" together, for example), then we see each language as a new manifestation of the unique ideas of a group of people, similar to a dialect in natural language. There are new idioms and expected constructions. The capability of expression remains constant, more or less, but the minor connotative differences are many and varied.
All this is to say: Python is not "should destroying". It's just another way of looking at the same system. One from which we can learn things when we seek to design new programming languages, certainly, but I don't think its idiosyncrasies inherently make it so bad as many portray it.