[SciPy-Dev] scipy documentation

Fri Mar 26 13:55:28 EDT 2021

Hi Pamphile,

thanks for sharing your opinion.

As I am new to this mailing list I am a bit unsure, whether discussing 
such a general and versatile topic via the mailing list in detail would 
be appropriate - but I at least want to be as polite as to answer you 
shortly.

Freedom is constituted by the freedom to choose, if you ask me.

Python has been a very free - e.g. multi paradigm - language from the 
start. Providing separated 'stable' (2.7.*) and 'experimental' (3.*) 
branches was a reflection of said spirit of freedom (and also a very 
clever thing). If you ask me, this spirit was a major reason for Python 
gaining popularity fast.

I do not deem it clever to now try to FORCE people to switch to Python 
3.* by taking away the possibility of installing 2.7.*. People do not 
like being infantilised, that could backfire; there are other nice 
languages out there... I e.g. am fluent in C# and WinForms as well - no 
problem to develop FRONTends with that instead in the future :) In this 
case SciPy naturally won't be part of it anymore.

By the way, I am not in the least sure, that "almost everyone has 
already migrated" and "the only notable example" is JPMorgan. We also 
would have to discuss, what "notable" is, first. Furthermore Windows has 
proven to be very downward compatible over decades. Your arguments might 
fit well into the scientific linux sphere or in the area of BACKends.

Blythooon is just an offer. No developer is forced to use it. Listing 
Blythooon is not the same as recommending Blythooon; it is possible to 
list something with a hint that you do not recommend the usage of Python 
2.7.* based systems anymore. That would let the people the freedom of 
choice.

But as I said, I am quite fine with the decision not to list Blythooon, 
it is your decision.

Best Regards
Dominik

On 2021-03-26 13:59, Pamphile Roy wrote:
> Hi Dominik,
> 
> Thanks for explaining the rationale behind Blythooon.
> 
> I am afraid that Python 2 is and must not be promoted any more, in any 
> way.
> Being out of support is really concerning, it means that it will not
> officially get security fixes, platform-specific fixes, etc.
> From now on, any new version of windows, or any other OS, could
> potentially break Python 2.
> 
> Almost everyone has already migrated to Python 3, production code
> included. The only notable example I know is JPMorgan which is still
> transitioning.
> But even they expect the migration to be complete this year. In the
> scientific sphere (at least all the communities I know), everyone has
> or is moving away from Python 2.
> You can have a look at all surveys showing how Python 2 is dying. For
> instance, https://www.jetbrains.com/lp/python-developers-survey-2020/.
> 
> In my opinion, we should call it a day for Python 2 once and for all.
> 
> Cheers,
> Pamphile
> 
> 
> 
>> On 26.03.2021, at 13:24, scipy at blackward.eu wrote:
>> 
>> Hi Ralf,
>> 
>> 
>> thank you very much for having a quick look into Blythooon and thanks 
>> for your fast and kind response!
>> 
>> NumPy 1.16.6 is the very last version supporting Python 2.7.* - as far 
>> as I know. That is the reason why it is used by Blythooon.
>> 
>> Blythooon packages the latest versions known to fit together around:
>> 
>> Python 2.7.18
>> PyQtGraph 0.10.0
>> PySide 1.2.2
>> 
>> Although the current dogma might be "switch to Python 3.*", the 
>> belonging ecosystem is not suitable for production quality software 
>> yet, at least if you ask me. This has several reasons. The primary 
>> reason is the vast number of minor and major incompatibilities between 
>> the different subversions and related packages and the risk which 
>> comes with each updating due to that.
>> 
>> Blythooon is thought for production (industry) quality, scientific 
>> applications with GUI. The purpose of Blythooon is not to provide the 
>> newest packages out there or an ecosystem which is permanently 
>> updating. The opposite is the case, it provides a "freezed" 
>> compilation. It is all about stability and compatibility. All 
>> Blythooon installations are 100% compatible.
>> 
>> You e.g. can check out the combination PyQtGraph 0.11.* with PySide 
>> 1.2.* - you then will detect, that this combination has severe 
>> incompatibility issues. It is the same with PyQtGraph 0.10.0 and 
>> PysSide 1.2.4 and so on. (You can easily try that out using the test 
>> program bundled with blythooon - the plots will not plot at all or not 
>> plot properly in said cases)
>> 
>> People using Blythooon do not need to resolve these issues on their 
>> own, they just get a working compilation - comprising a fitting SciPy 
>> and NumPy. And you might have noticed, that WinPython for example does 
>> not seem to support Python 2.7.* environments anymore. So, Blythooon 
>> fills a huge gap.
>> 
>> By the way, as long as the current Blythooon version is working 
>> without errors, there will not be (the need for) another release. That 
>> might somehow be the nature of a freezed compilation... It is somehow 
>> a snapshot of time.
>> 
>> But naturally I accept your decision not to promote it on the SciPy 
>> website yet. I just wanted to let you know the idea behind 
>> Blythooon...
>> 
>> Keep on folks, I dearly enjoy SciPy, you are doing a great work!
>> 
>> 
>> Best Regards
>> Dominik
>> 
>> 
>> 
>> 
>> 
>> 
>> 
>> 
>> 
>> 
>> 
>> On 2021-03-26 09:49, Ralf Gommers wrote:
>>> On Fri, Mar 26, 2021 at 12:17 AM <scipy at blackward.eu> wrote:
>>>> Howdy Folks,
>>>> I would like to suggest adding "Blythooon" (in contrast to
>>>> "WinPython"
>>>> this supports Python 2.7) to the list of "Scientific Python
>>>> Distributions" listed here:
>>>> https://www.scipy.org/install.html
>>>> Blythooon can be found here:
>>>> https://pypi.org/project/blythooon/
>>>> and the belonging installation step-by-step-guide video here:
>>>> https://www.youtube.com/channel/UCOE8xqYS_2azFsFjBVEwVMg
>>>> May I ask - how can I do that best? Thanks in advance and
>>> Hi Dominik, thanks for the suggestion. I had a quick look and
>>> Blythooon has only had a single release, last month, and ships numpy
>>> 1.16.6 only. That does not look like something mature enough that 
>>> we'd
>>> want to recommend it to the average SciPy user yet.
>>> Cheers,
>>> Ralf
>>> _______________________________________________
>>> SciPy-Dev mailing list
>>> SciPy-Dev at python.org
>>> https://mail.python.org/mailman/listinfo/scipy-dev
>> _______________________________________________
>> SciPy-Dev mailing list
>> SciPy-Dev at python.org
>> https://mail.python.org/mailman/listinfo/scipy-dev
> 
> _______________________________________________
> SciPy-Dev mailing list
> SciPy-Dev at python.org
> https://mail.python.org/mailman/listinfo/scipy-dev