[thehardsphere]

For those of you now breaking out the Internet Archive to prove the note existed

> "Yes, that is kind of funny way of saying that there's no reason why Python 2 and Python 3 can't coexist other than the Python project's incompetence and arrogance. Obviously it's theoretically possible to run Python 2 in Python 3, but until they do it then they have decided to say that Python 3 cannot run one other Turing complete language so logically Python 3 is not Turing complete."

Even though he says it's a "funny" way to say something, his explanation of said "joke" is not a correct description of what Turing completeness is. There is no "gag" here; Zed didn't understand it, and when people on the internet corrected him, he decided to play it off as a gag.

Edit: Just to make it explicit for those of you reading who don't get it, "Turing completeness" doesn't mean "I can run a program written in X in Y's runtime." It means that a language is capable of expressing all operations a Turing machine can do, which means that you can re-write a program in X to a program in Y if both X and Y are Turing complete. You can obviously re-write a Python 2 program to be a Python 3 program, so both of those languages are Turing complete.

The property he describes later, of the JVM and the CLR supporting multiple languages has absolutely nothing to do with Turing completeness. Lisp and Javascript are both Turing complete languages, but the fact that you can't run Lisp on Node.js doesn't mean you can't re-write a Lisp program in Javascript. The fact that he's equating being able to run many languages on a single runtime and Turing compleness means he doesn't understand what he's talking about.

[webmaven]

>> Obviously it's theoretically possible to run Python 2 in Python 3, but until they do it then they have decided to say that Python 3 cannot run one other Turing complete language so logically Python 3 is not Turing complete."

> "Turing completeness" doesn't mean "I can run a program written in X in Y's runtime."

That's not quite what he meant. He meant that Turing completeness of both languages implies that an interpreter for Python 2 can be written in Python 3,and that if the Python maintainers say it isn't possible, etc. etc.

Which is technically true, except that it is completely obvious that the performance of Python 2 under that arrangement would so totally suck that no sane developer would ever take that approach (and I can believe that the Python maintainers just said "that won't work" or some such), and that is what he apparently didn't understand (or is now claiming he pretended not to understand for comedic effect).

[lmkg]

> "Turing completeness" doesn't mean "I can run a program written in X in Y's runtime."

It's not the definition of Turing Complete, but it is a provable property of Turing-Completeness. If Y is Turing-Complete, then you can use Y to write an interpreter for X. Then the issue is reduced to arguing about what the word "in" means in the phrase "run Python 2 in Python 3."

[thehardsphere]

Actually it's reduced to the words "in Python 3" because the conflation between the Python 3.x language and the CPython 3.x runtime is something that is prevalent throughout the document.

Just because the maintainers of CPython 3 haven't included a Python 2 interpreter doesn't mean that Python 3 is not Turing complete. Their choice not to do that has nothing to do with the Turing completeness of anything.

[TuringTest]

> Just because the maintainers of CPython 3 haven't included a Python 2 interpreter doesn't mean that Python 3 is not Turing complete. Their choice not to do that has nothing to do with the Turing completeness of anything.

Which is precisely the point that the original version of the essay was making. The joke was on those insisting that it couldn't possibly work, and the punchline was telling them "oh, so what you're saying is that your language is not Turing complete?"

Ok, maybe that doesn't follow the format for a classic joke, but it's certainly a humorous and sarcastic remark.

[joshuamorton]

Except that the original statement isn't really that you can't write a python2 interpreter in python3. In fact, I'm quite sure that those exist somewhere on github, but that there wasn't some kind of transparent interop between py2 and py3.

That is, why can't the cPython3 implementation of the python3 language also execute python2 code? To which I think the simplest response is "why can't the gcc implementation of the C++ language spec link arm-compiled C++ and x86-compiled C++?"

[TuringTest]

> Except that the original statement isn't really that you can't write a python2 interpreter in python3.

That's why Z thinks of this as a joke, not a straight-face assertion.

[joshuamorton]

I don't know what this statement is supposed to mean, could you rephrase or elaborate?

[derekp7]

> but the fact that you can't run Lisp on Node.js

But you can write a Lisp interpreter in Node.js, because they are both Turing complete. In fact, I believe I've seen tutorials written up on how to write a Lisp in Javascript.

[thehardsphere]

Let me clarify further.

The fact that Node.js, the software package that some people use to execute Javascript, does not include in it is default distribution a way to execute Lisp programs, says absolutely nothing about the Turing completeness of the Javascript programming language.

What those of you repeating this interpeter thing are missing is that there's a four term fallacy being employed here. Zed is trying to make an argument of the form:

If A, then B If B, then C Not C, therefore Not B

But in reality his argument is:

If A then B If B, then C Not D, therefore not B

[TuringTest]

No, the argument being made was:

If A, then B.

If B, then C.

Some people insist that not C, therefore they are unknowingly implying not B.

But you have several comments dismantling an strawman version of the real argument, so it looks like you really didn't understand Z's intention.

His comment about people "failing to [in this case, understand] that note even when they do elaborate responses to [his] writing" looks dead-on.

[thehardsphere]

And who are these "some people"? I don't see citations here identifying anybody.

Pro-Tip: They're straw men. Even if you find examples of people asserting that after the fact, the undentified parties asserting "Python 3 cannot run Python 2" in the way that Zed claims he's responding to, are only mentioned in this rant to make Zed look smarter than he is.

[TuringTest]

You don't need real people to build up a joke.

[eastWestMath]

In other words:

http://i.imgur.com/pLOrJ.jpg

[sfilipov]

Agreed, it's not a problem of theoretical possibility. It's just that the maintainers' time can be better spent improving Python 3 than maintaining a Python 2 compatible runtime.

[srssays]

> You can obviously re-write a Python 2 program to be a Python 3 program, so both of those languages are Turing complete.

This is true, but the converse isn't true. e.g. Python and brainfuck are both turing complete, but Python has interfaces around a hell of a lot more syscalls than Brainfuck which only has operations for reading and writing from stdio. You can't write a Python implementation in brainfuck, it is impossible.

[thehardsphere]

No, that's just more work, because you have to write the required syscall interfaces for Brainfuck. There's nothing about Brainfuck that makes that impossible.

There's a difference between "impossible" and "really difficult giant waste of time" here when it comes to discussing Turing completeness.

[vurpo]

Actually, how would you do a syscall in Brainfuck? Its only way to interact with the outside world is through "input byte" and "output byte", and that does not give you the ability to do arbitrary syscalls. As far as I can tell, arbitrary syscalls are impossible to do in Brainfuck (which is why that one person made a fork of Brainfuck with added syscall support).

Of course, syscalls are not related to Turing-completeness in any way. Turing-completeness means that the language can express any computation which is possible to do on a Turing machine. Turing machine computations can not have any side effects, and syscalls are a kind of side effect. Therefore, having syscalls is not necessary for a language to be Turing-complete.