On 09/18/2014 08:26 PM, Paul Moore wrote:

> From this example, I'd like to see the following improvements to the process:
> 1. Somewhere I can go to find useful modules, that's better than Google.
> 2. Someone else choosing the "best option" - I don't want to evaluate

Well judgement is always required. As an example: search async version of the famous
Requests library on PyPi doesn't give beginners any immediate obvious choices. 

If we based off just popularity count, we will mislead users. The description ought to be
more descriptive, but how descriptive? Can author cover all the possible keywords?

This is why, naturally, most people use a search engine and eventually end up
either on stackoverflow or some blog post written by me \o/.

One idea, similar to npm or gem store is commentary. It can become spam so how about,
we can enable tagging and suggestion? What if people can suggest their workflow in a
more obvious way than a blog post?

If anyone have the time and interest, maybe scan each PyPI package and analyze how a
package on PyPI is being used by other packages. We can even go as fa
as scanning public repositories should anyone feel the urge to do that, privately.
That, however, is a M-B dollar industry, known as search.

Though like App Store, Play Store, gem store, whatever, the number of useful comments
and the number of participants can vary. Thus, the improvement is may be suboptimal.

On Fri, Sep 19, 2014 at 1:39 AM, John Wong <gokoproject@gmail.com> wrote:
I like stdlib++, but I also want to say it should remain as non-Python-dev endorsement
kind of thing. As a user and a library developer, I see pros and cons.

Official endorsement can lead to people to abandon whatever they are working
or make them feel excluded or unappreciated, which is not a very positive
thing to do.

On the other hand, there may only be a very limited number of stdlib replacement
people can vouch for easily.

For HTTP request, it is obvious that at the moment, Requests is the most widely
used library in modern Python codebase in the past several years.

For asyn-networking and asyn-task, Twisted and Tornado
are probably the best.
You might celery in the asyn-task just because many people who choose to
run asyn task will end up queuing.

For cryptography and security, what do you suggest? There are APIs cryptography
doesn't have yet but pycrypto or M2Crypto does. It isn't that neither are bad library,
I've used pycrypto and I am careful with using the API, but do we endorse cryptography
over the other two (the latter one is pretty dead based on its commit activity, I don't know).