[Scons-dev] Atlassian, BitBucket, Mercurial

[Dirk Bächle]

Hi again,

On 23.09.2015 13:45, Russel Winder wrote:
> On Wed, 2015-09-23 at 07:08 -0400, Gary Oberbrunner wrote:
>> Could be; pretty much _every_ project I can think of is using git
>> now.  I
>> have to twist my brain around every time I work on SCons. :-)
>

here are my 2cents regarding changing our VCS or project hosting:

Let's. Not. Switch.

You ask why? Because I'm happy. I'm so happy that---after a seemingly never-ending series of events, with project managers leaving, 
switching from SVN to hg, migrating our Wiki to Bitbucket, and so on---we've finally reached a state where I can do what I like 
most: coding.

It's tempting to think: yeah, let's just import the current repo into git, push it to Github, and we're good to go again. But I 
count myself to be a member of the core developer team. So, in the case of switching to another VCS I'd feel obliged to work on 
setting up and properly documenting the new workflows in the Wiki. I'd also feel obliged to answer questions from newcomers, or 
people getting confused over the VCS change.
Which is both fine per se, but it would keep me from doing what I like most (and am good at, to a certain extent): coding.

But anybody out there has a good chance of convincing me that switching is good for the project overall, by (a priori to the actual 
switch) either:

   - providing a full replacement for the current descriptions of our general development workflow, and about howto creating pull 
requests in particular, for our Wiki and volunteering to answer all upcoming questions about it, or
   - jumping on a large bug or issue (like the python3 branch, for example), setting up his preferred git-hg bridge, working on 
stuff, and wowing me and everybody else with the way he makes impressive progress in a very short time. ;)

> Which is sad because Git still has so much wrong with it, and Mercurial
> has so much going for it. However, the social and commercial pressure
> seems inexorable that Git shall be the monopoly.
>

I don't feel any pressure in this direction. (see also below)

> I think there possible needs to be a new debate about whether to switch
> to Git and GitHub.

I don't think so.

When the decision was made to move from Subversion
> on Tigris to Mercurial on BitBucket, I had some issues/doubts, but the
> final decision seemed fine and not a problem. However time has moved on
> and we are are now in an era where to get any traction as a FOSS
> project you appear to have to use GitHub.
>

Can you elaborate a bit on this conclusion? I fail to see the connection between "Project is at Github" and "Project is successful" 
as a direct consequence. How exactly do you get from one to the other?
Isn't it more the fact that, since a lot of projects are at Github (for whatever reasons), there's a high probability that a 
successful project is at Github? Have you checked for the opposite side of the scale? My bet would be that the worst doing projects 
are at Github too...because almost everybody is there!

I always thought that to get any traction and have success, you need to have a good product/project. Why don't we continue working 
on that in the first place?
We have a full roadmap and TODO list, and last time I checked there were 1500+ unresolved issues in our Tigris tracker. ;)

> OK, I realize it is all "kool kids", "hipsters", and "people who say
> awesome evey sentence", but the issue is not one of what is best as a
> technology, it is what is best as a way of gaining traction.

See above for my notion of "traction". ;)

I
> appreciate Mercurial is written in Python (mostly) and Python users
> Mercurial, but they have PSF. SCons is just another ancient build
> system trying to exist as a FOSS project.
>

This feels to me like we should accept getting dictated "from the outside" what to use in our project. I think that SCons as a 
(Python!) project has a good standing and can afford to "flex its muscles" when needed. You say that Mercurial gets left by many 
projects, and we should do the same? Then who's going to stay and ensure that Mercurial doesn't die (sorry, couldn't think of a less 
dramatic word right now)? Isn't this exactly our job, as part of the Python community?

> I propose therefore that we switch from Mercurial/BitBucket to
> Git/GitHub and that we start a new GitHub issue tracker (even though it
> is shite really). Issues from Tigris can be moved over by hand as
> needed.

I don't migrate 3000+ issues by hand. Funny thing is, only yesterday I continued working on my converter script for importing all 
Tigris bugs into a Roundup tracker instance. There are just a few minor unresolved problems with it, so if you think a different bug 
tracker "cuts the mustard", please join and help me with it.

This then gives access to Travis, CodeShip, etc. The point here
> is not that they are reliable CIs, it is that they are ways of
> promoting and marketing the project in the Git-sphere.
>
> Given Buck, Bazel,… SCons needs to do something to make a new presence
> or simply drift into the dustbin of history.
>

This is the first time ever that I hear about Buck and Bazel...which proves to me that being at Github is no guarantee to getting 
noticed as a software project.

Best regards,

Dirk