Still digesting the OVH fire in Strasbourg

Monday, March 15, 2021

Hannes is working to get the mail servers at lino-framework.org and saffre-rumma.net up and running again. Yes, we didn’t have a backup plan for these servers because they don’t hold any vital data, all knowledge on them is documented under Setting up a mail server. But it is still quite some work to actually apply it again. I don’t complain. It is a challenge that will help us to become better.

We need a better nickname for our new backup server because its current nickname “hetzner2” is not a name we can show to external partners. It must bomething neutral and easy to remember. It won’t be “famous”, because only team members and business partners will talk about it. I choose bombur. Here is my speech at the naming ceremony:

Bombur is one of the dwarves who were companions to Bilbo Baggins on the quest to reclaim Lonely Mountain from the dragon Smaug, along with his brother Bofur and cousin Bifur. These three were to become something like an honour-guard to Thorin. (source). They were not descended from the Line of Durin, but rather the line of Moria. Considered working class, Bofur was a miner like his parents before him. Bombur eventually does prove to be a loyal person (source).

I restored my developer blog (https://luc.lino-framework.org) and the Community Guide (https://community.lino-framework.org/). Here is my cheat sheet for adding a new static site:

cd /etc/nginx/sites-available
sudo cp lino-framework.org.conf luc.lino-framework.org.conf
sudo nano luc.lino-framework.org.conf
sudo ln -s /etc/nginx/sites-available/luc.lino-framework.org.conf ../sites-enabled/
sudo service nginx restart
sudo certbot --nginx -d luc.lino-framework.org

I am still hesitating whether we should restore the pre-Strasbourg URL system where we had a separate subdomain for every little doctree. We might have a single subdomain docs.lino-framework.org where all these “minor” doctrees (atelier, etgen, getlino, cosi, tera, noi, …) would be. One disadvantage is that we probably need to add yet another doctree: the one that will be served at the subdomain docs.lino-framework.org itself. Why not. The “Lino Documentation Centre”, nickname dc.

But the first precedent against the idea in the previous paragraph was quick to come: it was https://eidreader.lino-framework.org, for which I “inadvertently” added a subdomain and a certbot certificate. I guess that this will continue with the other minor doctrees (algus, cosi, noi, avanti, tera, presto, welfare, …)

What about links like https://de.welfare.lino-framework.org and https://fr.welfare.lino-framework.org, which have been published in miscellaneous channels and will probably hang around forever? AFAIK this can be fixed by configuring permanent redirects.

I made another –probably last– commit to the GitHub repository of the Book (https://github.com/lino-framework/book). This repository is a bit dangerous now because people may not realize that it’s no longer being used. So I added the following sentence to the beginning of the README.rst file:

Warning: this copy of the Lino Book is obsolete. New project homepage is https://gitlab.com/lino-framework/book. See https://www.lino-framework.org/team/gh2gl.html for details.

I discovered a gaping hole in the documentation (it seems difficult to find contact information), and that we need a “Lino Contributor License Agreement” to be signed by every contributor. I opened #4022 for this. What a Pandora’s box! It feels like an absurd waste of time that I need to care about legal issues, and that I must ask a signature from a volunteer who wants to contributes a code change. But yes, I know that the current copyright legislation requires it, I even knew it before. The incident in Strasbourg reminded me that shit happens. So I need to get my hands dirty and start to care about these things.

Two standard files CHANGELOG.rst and CONTRIBUTING.rst should be in every repository. For the Lino framework we will use .rst suffix. Other projects, as it seems, use either .md or no suffix at all.

As a personal side effect of the fire in Strasbourg, I didn’t have enough sleep for several days, and I try to slow down until my sleep is back to normal. Better a few slow days than a burnout!

The doctest docs/specs/noi/db.rst reports social_django.UserSocialAuth to have 5 database fields when running on GitLab, but 7 fields when running on my computer.

Here is the model definition on my computer:

class AbstractUserSocialAuth(models.Model, DjangoUserMixin):
    """Abstract Social Auth association model"""
    user = models.ForeignKey(USER_MODEL, related_name='social_auth',
                             on_delete=models.CASCADE)
    provider = models.CharField(max_length=32)
    uid = models.CharField(max_length=UID_LENGTH, db_index=True)
    extra_data = JSONField()
    created = models.DateTimeField(auto_now_add=True)
    modified = models.DateTimeField(auto_now=True)
    ...

That is, 6 fields, plus one for the primary key. 7 Fields. So what is happening on GitLab? We use te latest development snapshot (in the book’s requirements-include.txt file):

social-auth-app-django@git+https://github.com/python-social-auth/social-app-django

Aha, but the setup_info.py file of the book specified yet another dependency to social-auth-app-django==3.1.0.