Implement the instance picker

[thibaultamartin]

Fixes #1570

This work is largely based on Austin Huang's great work on servers.joinmatrix.org

TODO:

• Figure out the data model for instances
• Implement the dialog view for the instance details
• Import all the instances data
• Update the instances data with a daily CI job

[thibaultamartin]

Data Model for instances

The two options I have in mind are:

1. A single toml file that any interested party can easily consume.
2. One .md file per instance, with structure information in the frontmatter.

Single toml file approach

Pros:

• Very easy to consume for third parties (they have a single file to fetch and parse)
• Works well if we don't have a dedicated page per instance for the details but a popover

Cons:

• Probably makes it much more difficult to have a page per instance
• Popovers work well with javascript, but are difficult to do (and most of the time not accessible or a mess) without

One markdown file per instance

Pros:

• Makes it easy to create a dedicated page per instance for the details

Cons:

• More difficult to browse and parse for third parties

---

I'm going with the single toml file approach, which seems to be the easiest to consume for third party clients for now. We can refactor later.

[jplatte]

More difficult to browse and parse for third parties

Zola is not limited to HTML as an output format. It should be possible to write some Tera code to expose the data stored in frontmatter of multiple files as one JSON file, for example.

[thibaultamartin]

Is it possible?! That would entirely save the day.

If you have pointers to the right docs that would be very helpful

[jplatte]

Uhh, now I'm wondering if I am imagining things 😅
I know it used to work in a static site generator I used at some point, might have been Jekyll where every content file was processed as long as it contained frontmatter? Certainly not seeing anything in docs – it's not even documented that you can use .html files instead of .md? I'm pretty damn sure Zola supports that though.

[thibaultamartin]

Updating the CI data

The only bits that should be updated automatically are the software used and the version it relies on.
All of this information can be retrieved via the version endpoint.

Typically for ergaster.org:

thib@MacBook ~ % curl -s https://ergaster.org/.well-known/matrix/server | jq '."m.server"'
"matrix.ergaster.org:443"
thib@MacBook ~ % curl -s https://matrix.ergaster.org/_matrix/federation/v1/version | jq '.server.name'
"Synapse"
thib@MacBook ~ % curl -s https://matrix.ergaster.org/_matrix/federation/v1/version | jq '.server.version'
"1.91.1"

Then either the toml file needs to be parsed and updated, or the markdown files need to be parsed and updated depending on the data model chosen

[thibaultamartin]

Current state of things. It's not responsive yet.

[cloudflare-pages]

Deploying matrix-website with    Cloudflare Pages

Latest commit:	5f6b46a
Status:	✅  Deploy successful!
Preview URL:	https://94ed3ee5.matrix-website.pages.dev
Branch Preview URL:	https://instance-picker.matrix-website.pages.dev

View logs