This looks pretty amazing!

I think the gossip protocol might hold it back, as does the byte limit. Part of what made the early internet amazing was the longer content. Maybe a 10k limit would allow for a little more freedom?

A distributed hash table(Perhaps like OpenDHT, of which I'm a big fan because of the standardized HTTP gateway protocol) could store far more than 10 million of these.

Current DHTs might not be appropriate, as the churn is rather high and they are based on the idea of republishing every 15 minutes or so, but maybe a more persistent DHT can be developed, one that discourages ephemeral nodes in favor of a smaller number of higher bandwidth ones.

I think compression should also be part of the protocol, just because efficiency matters with this kind of non-commercial stuff. RAM and Disk are still rather pricy on a cloud server, for the kind of thing individuals do.

This also seems like it could be a somewhat nice use case for Microsoft Adaptive Card templates rather than HTML.

It's complicated, but if you have a very limited amount of data to work with, it might be nice to not worry about presentation at all, just JSON with a $schema key telling what format it is, allowing the client to use it's preloaded template to render it.

The early web seemed to be a lot more content-centric, leaving more of the rendering and visual distinctions to the browser.

Another nice thing is they are semantic and machine readable, in case someone wants to publish some weather station data that way.

Command line clients might also be a lot easier, using cards made of standard templates instead of freeform HTML.

But, sometimes the style is critical for the content, so Adaptive Cards could probably only be a replacement, not an alternative.

One really amazing thing from the web I remember was the web ring. Having boards be more structured, with at least some metadata, might allow for some kind of webring-like graph, where people could link to other boards in a really simple, semantic way, that would allow for crawling and discovery of lists of related sites.