> You're effectively prohibited from running a web browser or telephone from 10 years ago because of planned obsolesce.

But, but, but it is for your own security. /s

Those Python 2 programs can reasonably easily be repaired by porting them to Python 3.

This is exactly the thing.

If the difference between python 2 and 3 is so small that a program can be easily ported, then WHY did there need to be a difference at all?

Usually change happens because the old way was worse and the new way is better. For each individual porting change, you would have to look up the appropriate PEPs to discover why the change was made.

That's the claimed reasoning, yes. Is that actually true though?

It may be from a pure economic perspective focused on a business that can socialize externalities. Sometimes new developments reduces development time. On the other hand, the externalities of that is that a modern computer performs worse running equivalent tasks today in the browser than a native app 15 years ago on hardware from that time. Lucky for businesses that they aren't held responsible for the ewaste they helped produce.

Web development isn't mandatory, you can still have a native app in 2023.

The point is that the lack of businesses being made responsible for the externalities they cause disincentives native apps, not that it is impossible.

If 90% of apps necessary for normal life are only available as poor web versions it drives obsolescence no matter what the remaining 10% do.

Why is web app a bad thing though? Most of the apps do not require much performance and the web version can also be pretty fast, at least sufficient for users. I think the same author writing that crappy web app will produce worse native apps if forced to do so.

Web apps is better than native as they are automatically running the latest version, platform independent (windows, mac, linux, android, ios), in some sense more secure. Instead of asking people to write more native apps, I think we should rather ask them to write more quality web apps if they don't need native capabilities.

I started writing yet another rant about how most HN users don't care about the rest of the planet, but then thought better of it. Long story short: way more people on Earth are using cheap (and thus low-power) hardware then you probably think. And it isn't necessarily old.

Try running eg. Libre Office and gdocs on a 10+ year old laptop. The difference is night and day. And that is a web app which should not be particularly demanding (essentially text entry).

No, google docs is not just text entry, it is much more complicated than that due to synchronization and they did not try to optimize it well IMO. Text entry, even with highlighting and rich text, can be fast with web app, see VSCode and markdown applications for example.

As an example, see Atom vs VSCode. We thought building a code editor with electron will lead to something slow like Atom, but m$ later proved that it is possible to build something fast. Perhaps not as simple as native, but it can be done.

> That's the claimed reasoning, yes. Is that actually true though?

yes.

one of the changes (and in my experience, the most problematic change) was the removal of the keyword "print"

it was a poor choice at the start to treat "print" like a keyword. it is a simple, breaking change to remove it, easily automated (ps: just add brackets), and now the language is more consistent and flexible.

The only reason to keep it is to not "touch" things. but it is easily mitigated, so that isnt a good enough reason imo. I much prefer a consistent language.

otherwise I'd stick with PHP.

Is it easilty mitigated though?

By a python programmer, sure.

By another programmer, ok.

By your grandpa that just wants to run that python script to organize his stamp collection that has worked for decades? No.

For something as popular as python there will be thousands such people. Stuff that will simply NOT get fixed.

But that cost is easily ignored by the original developers.

There's no point positing a less likely reason. Python wanted to incorporate Unicode fully. There was no way to do that that they came up with that didn't break a lot of software. You can read the documentation and see the discussion. It's not companies; it's a few people in Python leadership deciding.

WikidPad had (??) a fairly large user base. The Win32 executable last released still works with my personal notes from 2008 onward.

There was a breaking change in Python, followed by a breaking change in WX_Windows, so while you can, in theory, run WikidPad under Linux today, few of the dialog boxes work, effectively disabling the program. 8(

Python2-3 transition COULD have been made easier, but it wasn't. The community places far too little emphasis on backwards compatibility.

This is because there's a myth: pushing a free upgrade across the internet is zero cost

This is not true, and can't be true. The cost of manufacturing software has effectively fallen to zero. This is not the same as the cost of implementation.

--- Tangent/Analogy ---

When you get a large machine tool, there are physical installation requirements that can delay its use by up to a year. Sometimes the floor needs to be removed and replaced with a better foundation. Power, cooling, air, vacuum, networking and other connections need to be made. These may require updates to infrastructure.

Once sited, machines have to be precision leveled, and there is always some form of break in period as differences are found and accounted for in their use.

Users have to be trained to allow them to work with/around any UI or performance changes that might effect production.

---

In a very similar fashion, software requires infrastructure (dependencies, operating system support, bug fixes, user training) and new executables carry new bugs to be worked around, performance changes, etc.

Security fixes should strive to minimize changes, but to me it is entirely unacceptable that they are necessary, especially at the rate currently required.

Security should be the job of the Operating System, not the application, network, users, especially NOT the users.

Software security updates should happen about as often as automobile recalls. They should be costly to the manufacturer and infrequent.

True, but not a soul is trying to steal your credit card via BC-232.

This is the entire point of setting up an environment, to call up a known good dependency for this use case. None of those python 2 programs went away. They are still there, waiting for you to run them with python 2.

> Uppity programmers calling themselves software "engineers" act in almost totally the opposite manner of a true engineering discipline. They hold "legacy" in contempt, and break things in the name of "progress" or "safety" at odds with their stated goals. They ignore the lessons of history, and keep re-inventing the same things every decade or two.

The reason for that is that software is much easier to change than hardware. You cannot expect every person in the world to update their radio equipment every couple years - it's too expensive - but it's somewhat reasonable to expect people to install new software, especially if it can be done for free over the internet.

It's just economics at play.

That's not true most manufacters can change the machine yearly but try to change sap for oracle, or your custom solution for sap, and you are spending million of dollars in multi years projects, I'm not expecting free and cheap software to end , but having a game than breaks in 10 years isn't problematic, but pushing delicate software for no reason, that could break in 5 years, becouse is impossible to the customer to know the dependencies of the program, or how robust is under stress test and cybersecurity, means we lack engeniering discipline not economics reasons.

I think this is purely economic reasons, engineers working on that project is unlikely the people suggesting such a project...

Software is still easier and cheaper to change than what you'd spend for changing hardware, data and users. It's every time a multi-year project for big programs because of data and accustoming users for their habits and needs.

>it can be done for free over the internet.

The cost to manufacture an executable is effectively zero

The cost to integrate an executable is NEVER zero, there are always costs.

Windows is far, FAR less reliable on Wednesday morning because of "Patch Tuesday", this is a substantial and real cost.

The same is true for any system that implements "security patches" and feature "upgrades".

There are programs, protocols, ecosystems and programmers that favor stability. However seems to be that the closer you get to the user, the less stable and simple things are. It's ironic or maybe even hypocritical of us programmers to prefer stable tools for ourselves, but then move around and break stuff for our users every few months.

Stability is really something that we need to get a lot better at as programmers. Perhaps accesibility laws/guidelines should limit UI changes for automatic/required updates - you can always have users opt into your shiny newfangled UI if you want, but leave the old one instead of confusing your users.

After 10 years should farmers have to give away their land for free too?

Farmers pay property tax and use the land productively. Property tax is an incentive to use the land productively.

Farmers have to put in constant work to generate revenue. "IP" revenue too often entails create-once-earn-forever. However, "IP" often has a higher upfront investment attached to it, and often is not (as) "scalable", unlike farming.

What I am trying to say is that both things are very different in details where it matters, so I think this is a rather flawed analogy.

Maybe a closer analogy would be landowners who rent out land. Should a person who bought land 10 years ago be able to generate revenue from renting it to a farmer?

If land could be replicated with a click of a button with close to zero cost, then sure.

If you're implying that the farmer would still have an income then it's not a valid analogy. If I open sourced my software that I've been continuously been developing for more than 20 years my income would drop to almost zero, immediately.

Your changes made in the last ten years would still be your exclusive monopoly so your revenue would probably not change much if you really have been continously developing it.

But even if this would destroy your business model, that doesn't mean it would not be a good idea for society as a whole. Business models become obsolete all the time. If there is enough demand for your software, you will be able to fund the development of it even without might-as-well-be-perpetual exclusive control.

If it is imaginary land that they have thought up, yes.

Meaning you don't value the time that someone puts into building software as highly as the time a farmer puts into paying his mortgage.

Ten years is an eternity when it comes to software. But yes, software developers (and other people creating things covered under copyright) are overvalued compared to those creating physical things. And that valuation is created entirely due to IP law. How is that fair?

>And that valuation is created entirely due to IP law.

In that case, the value of land is created entirely by property law.

A farmer has the exclusive right to land from the existing fixed supply that the entire world has available to it. He didn't create it and no one else can use it. Whereas a software developer makes something new. Nobody else was deprived of anything pre-existing. Less justification for taking code from a developer than land from a farmer.

Farmers will keep generate revenue with their lands 10, 20 years later. But Microsoft won't generate revenue with Office 2003.. This is whole different.

> There are a wide range of programs that broke when Python 2 went away.

Python 2 still exists. Those programmes will break if you try and run them on Python 3, but that's not the same.

A few distros have stopped shipping it in a usable manner in their repos.

Indeed. Additionally, people can install it themselves, or run their programs through a Python 2 Docker image.

> Even fully open sources software can't be "repaired" when its dependencies break. There are a wide range of programs that broke when Python 2 went away.

Can't? Like physical objects can't be repared when the parts are no longer sold?

There is always a way to repair, just a matter of how difficult it is. Having the sourcee to the application itself is the first difficulty barrier. Source to dependencies would also be helpful, yes.

Btw, Python 2 did not go away unless you mean it went away from the default package archives of some Linux distributions - but even there it's not (yet) gone for all of them. The Python 2 sources are still the same as they have always been and while you might need older versions of a few more dependencies, if your really need to have Python 2 you can get it.

Python 2 didn't "go away", you can still use it without issues.

If you step outside of the myopic web space most well written native code (SYSV ABI) has a decent shelf life.