>> Many Schemers are also in gitlab nowadays.
>
> Not sure if running parallel hosting providers is worth the effort,
> especially in my plan for discussion to happen in pull request
> comments, to be scraped and archived in a single message batch to a
> mailing list when the request is resolved.  But the current obvious
> place to move if GitHub/Microsoft betrays us.

Either host is fine with me. As you say, it's Git so we can always move.

>>> using the rsync.net backup account
>>
>> Backing up there is fine.
>
> Indeed, but people other than me have to start doing it.

Backup concerns ought to be much the same no matter which Git host and
web server we use, right?

>> Yes, e.g. a shell script that runs rsync :)
>
> Turns out rsync.net also provides a git server!
>
> https://blog.kozubik.com/john_kozubik/2010/02/git-and-subversion-support-at-rsyncnet.html
>
> https://www.rsync.net/resources/howto/github.html

Those blog posts simply log in via ssh and use command line git to clone
repos as usual. That's nice, but can be done on any Linux server.

> Raw storage is expensive compared to dumb object storage like AWS's
> iconic S3, or Deep Glacier at 1 US$/TiB/month (!!!), but the use and
> support of it is really attuned to people like us.  It's overall a
> class act owned and run by one of us vs. a "suit", let alone VCs.

I meant rsync for uploading things to our web server (a very simple
task). For the source data, We should store everything important (that
cannot be easily re-generated using a script) in Git repos. There are
now so many places to push repos (e.g. GitHub and others can host
gigabyte-sized repos) that Git is effectively a standard file transfer
protocol and backup medium. And it keeps history in a standard way that
we already know how to use. rsync is nice, but it's error-prone for
anything complex especially when multiple people work on the same data.

Re: backup storage, there's also <https://www.tarsnap.com> which is a
hacker favorite. But I'd recommend Git for all our stuff.

For things that can be generated, we can use Scheme scripts, and for
servers, there's infrastructure-as-code (Ansible or the like) which
means our server configs can be tracked in git.