Re: GitLab CI setup file in scratch/tzz/gitlab

[Dmitry Gutov]

On 01.05.2017 1:19, Richard Stallman wrote:

>    > >    > Great. Only Gitlab requires a file in the master branch, out of the
>    > >    > three we're evaluating (the other two are Hydra and Buildbot).
>    > >
>    > > Are those SaaSS too?
>
>    > That's a trick question. Gitlab is not SaaSS.
>
> It is an honest and serious question.  However, I should
> have distinguishd more clearly between the software called GitLab
> and the service called GitLab.

It's an important distinction, yes. And I don't think the question of 
whether Buildbot and Hydra are "SaaSS too" is important for our 
purposes. The important question is whether we can run them for ourselves.

Somewhere, I'm sure some company (or several ones) provide them as paid 
services.

> The GitLab software is not SaaSS, because it's not a service.
> It is a program.  One version of that program is free.
> It is fine for us to run that program if it serves our purpose.

Naturally.

> To use a commercial service for this would be SaaSS, even
> if it is running _the exact same program_

We can call it SaaS, which is the industry-standard term. To use the 
disparaging variation, against a well-behaving company that distributes 
a fully-functional version of said software with a Free license, I 
wouldn't consider fair.

> SaaSS is not a question of which programs are running on the server.
> We can't verify which programs are running on anyone else's server.
> We can't change them either, even if they are free software.

That's also true, for many users, of the GNU hosted installation of 
Hydra. And it will remain true, for most Emacs developers and users, of 
the GitLab installation on the FSF premises, when it's set up.

And yet, I wouldn't be comfortable calling them SaaSS.

> Please see https://gnu.org/philosophy/who-does-that-server-really-serve.html
> for an explanation of the SaaSS issue.

Already read it, thanks.

> Perhaps the FSF should find volunteers to set up a server to do CI for
> GNU packages using the GitLab software.  Would that be useful?

That has always been the idea, yes. We are only using the GitLab service 
to evaluate the features of GitLab the program.

Like already mentioned, by the way, the contents of .gitlab-ci.yml, 
which is proposed for adding into the Emacs repository, are not going to 
change after that transition.