"More About Unicode in Python 2 and 3"

[Terry Reedy]

On 1/5/2014 8:16 PM, Ned Batchelder wrote:

> OK, let's see what we got from three core developers on this list:

To me, the following is a partly unfair summary.

> - Antoine dismissed the post as "a rant".

He called it a rant while acknowledging that there is a unsolved issue 
with transforms. Whether he was 'dismissing' it or not, I do not know. 
Antoine also noted that there does not seem to be anything new in this 
post that Armin has not said before. Without reading in detail, I had 
the same impression.

> - Terry took issue with three claims made, and ended with, "I suspect
> there are other basic errors, but I mostly quit reading at this point."

You are discouraged that I quit reading? How much sludge do you expect 
me to wade through? If Armin wants my attention (and I do not think he 
does), it is *his* responsibility to write in a readable manner.

But I read a bit more and found a 4th claim to 'take issue with' (to be 
polite):
"only about 3% of all Python developers using Python 3 properly"
with a link to
http://alexgaynor.net/2014/jan/03/pypi-download-statistics/
The download statistics say nothing about the percent of all Python 
developers using Python 3, let alone properly, and Alex Gaynor makes no 
such claim as Armin did.

I would not be surprised if a majority of Python users have never 
downloaded from pypi. What I do know from reading the catalog-sig (pypi) 
list for a couple of years is that there are commercial developers who 
use pypi heavily to update 1000s of installations and that they drive 
the development of the pypi infrastructure. I strongly suspect that they 
strongly skew the download statistics.

Dubious claim 5 is this: "For 97% of us, Python 2 is our beloved world 
for years to come". For Armin's narrow circle, that may be true, but I 
suspect that more than 3% of Python programmers have never written 
Python2 only code.

> - Serhiy made a sarcastic comment comparing Python 3's bytes/unicode
> handling with Python 2's int/str handling, implying that since int/str
> wasn't a problem, then bytes/unicode isn't either.

Serhiy's point was about the expectation of implicit conversion 
(int/str) versus (bytes/str) and the complaint about removal of implicit 
conversion. I suspect that part of his point is that if we never had 
implicit bytes/unicode conversion, it would not be expected.

-- 
Terry Jan Reedy