Re: synapse: minor changes.

[TinfoilSubmarine <TinfoilSubmarine@users.noreply.github.com>]

[-- Attachment #1: Type: text/plain, Size: 3304 bytes --]

New comment by TinfoilSubmarine on void-packages repository

https://github.com/void-linux/void-packages/pull/34676#issuecomment-1004190591

Comment:
> Syslog is not installed by default on void so you won't see the problems raised anywhere. If you think that installing socklog-void and knowing where vlogger logs to by default is easier than using `ps` and `pgrep runsvdir | xargs -I{} cat /proc/{}/cmdline` you should consider reading up on [runit](http://smarden.org/runit/). Which is the init system that void is using.

Not trying to strawman here, but by this logic, it seems like no service in Void should be shipping a `$service/log` since we can just dig logs out of `/proc`? I'm legitimately curious what constitutes a valid reason for adding a logging service then.

> That's a nice form of cargo-cult that begs for the statement that if you like how it's done in distribution X maybe you should switch to distribution X. If I remember correctly void is trying to provide packages as 'vanilla' as possible.

I wasn't really pointing to "how Debian does it" as much as "the authors are using this approach as a way to ship their own software so maybe we can learn from it", but you are right that this may be a too vendored approach for Void.

> If one seriously wants to set up a synapse homeserver, they have to install synapse, get valid ssl certificates, set up a reverse proxy, install postgresql and set it up, add well-known files for clients and federation, have a working e-mail setup, set up a TURN server and start running different worker processes to do load balancing. I don't think that copying sample config files so that the service starts without failing after a fresh install helps a beginner doing any of that. But it might have wrong defaults like the name of the homeserver (which makes the generated sqlite-db unusable if you want to switch the name later) and missing randomly generated secrets as already mentioned. It also introduced problems for people that were already having a working setup.

I agree that making a few changes to a default config to get the synapse service running is not representative of the amount of work needed to actually stand-up a feature-complete synapse from scratch. The original reason I attempted to have a default config in place was to remove the existing `INSTALL` telling the user what to do, since that vendors setup instructions/info. Indeed, there's a lot more that the user starting from scratch will need to do than just run:
```
sudo -u synapse python3 -m synapse.app.homeserver \
      --server-name my.domain.name \
      --config-path /etc/synapse/homeserver.yaml \
      --generate-config \
      --report-stats=yes
```
which is all that `INSTALL` was telling new users to do. But, including this in `INSTALL` may be more in line with upstream documentation, since the [upstream installation instructions](https://matrix-org.github.io/synapse/latest/setup/installation.html) don't have an explicit step telling the user to generate their config file except if following the PyPI/virtualenv instructions. And [this page on configuration](https://matrix-org.github.io/synapse/latest/usage/configuration/index.html) just says that the configuration should have been generated when you installed synapse.