I think that the ideas can also be combined well. I really like the idea
of the pull model, the data could then be automatically rendered to
Markdown and put into the QUIC WG Wiki. This could happen for example
every week by a Github Action.
Before I built QUIC Explorer, I also had a big Markdown file with
information about congestion control of QUIC implementations. As Lucas
already mentioned, Markdown is not structured and cannot be properly
filtered, processed or machine-read. The option of storing the data in
JSON and rendering it flexibly to a static website (or Markdown) quickly
seemed more suitable to me. I just pushed a short 30-line proof-of-work
script here
<https://github.com/MPK1/quic-explorer/blob/main/cli/create_markdown.py>
that produces Markdown that looks like this
<https://gist.github.com/MPK1/cffb5c6a56a75e3d604a8835e2d55c2b>.
I also just created a json file with some structure that could be
improved and used for a pull model
(https://gist.github.com/MPK1/8765fc25bb98d437491ebcc6c55d09a9).
Marcel
Am 20.07.24 um 01:14 schrieb Lucas Pardue:
On Fri, Jul 19, 2024, at 06:44, Christian Huitema wrote:
On 7/19/2024 12:42 AM, Marcel Kempf wrote:
> Hi Lucas,
>
> This is also the way this projected started: People asked me if
> something like it exists and after saying no again and again, I
realized
> the need for it, especially as the QUIC Implementations wiki page on
> Github (https://github.com/quicwg/base-drafts/wiki/Implementations) is
> also quite outdated and not really extensible.
>
> Considering the pull model: I have not yet thought about that, but
> that's a great idea, thank you! I already have some nice ideas what
> could be possible, so I opened an issue
> (https://github.com/MPK1/quic-explorer/issues/5) and will add some
> proposals soon.
If we have a defined format, cannot we use it in the QUIC WG Wiki?
There's options for sure, and I wouldn't want to rule any out at such
an early stage.
Speaking of only my own experiences. The QUIC wiki was vandalised at
some point in the recent past and I locked down permissions. I can't
recall how restrictive those are. Possibly too restrictive to let the
ecosystem continue to bloom. Furthermore, having to use markdown
doesn't seem particularly structured to me. JSON in markdown sounds yucky.
The model I personally like is for projects to define their own source
of truth that is owned by them and pulled from. This seems to work
pretty well for the interop runner tests.
As another source of inspiration, see
https://github.com/mdn/browser-compat-data, which contains browser
oriented machine readable data that is used by a variety of tools such
as MDN, caniuse and more.
Cheers
Lucas
-- Christian Huitema
