Open Stack

Thomas Goirand wrote:
> On 03/14/2016 03:28 PM, Davanum Srinivas wrote:
>> Ian,
>>
>> +1 to get rid of that dependency if possible.
>
> +1 for any action aiming toward removing *any* dependency.
>
> We don't have enough of such actions, and we have a way too many
> dependencies, with many duplicate functionalities too. Just to name a few:
> - pecan vs falcon
> - oslo.concurrency vs lockfile
> - nose vs testr vs pytest
> - pymemcache vs memcached
> - you-name-it...
>
> And this isn't even motivated by the fact I maintain lots of packages, I
> don't maintain httplib2 for example, so I'm not impacted much,
> especially by this kind of package that don't upgrade often.

Converging dependencies is a bit of a thankless debt reduction job: you 
have to push changes in a lot of projects, and those are rarely seen as 
a priority. It's a bit like pushing for Python 3 compatibility... you 
need to find someone caring enough about it to persist in pushing those 
changes, otherwise it just doesn't happen.

We could have a squad of "convergers" that would define a very small 
list of targets every cycle and push that through.

-- 
Thierry Carrez (ttx)