feature request: make a simple protocol / API to allow servers to self report, for indexing about devzat

[elendee]

Just self-reported, publicly available info I think.
So I'd think

• created date
• number active users (there is no "registered users" in devzat right... although you could track all-time logins?)
• name
• description
• owner handle (I'd make the owner type this explicitly during config so it's not an email etc by accident)
• sub-channels / topics available

[quackduck]

That’s an interesting idea!

[elendee]

yea even better.. I was thinking this would be a 3rd party listing, a website etc, but you could build it right into devzat core, as long as there was a registry server online for it to ping.

.. although.. that would imply switching SSH sessions, which is probably out of the scope of one devzat install to do, since devzat lives inside one SSH session.

You could run a "devzat hub server" that listed online servers and tunneled people between them, but that would be a lot of bandwidth for you and defeat the purpose of distributing it.. so yea probably some level of client initiative is needed here, where they browse the "hub" listing by themselves and choose between them

[l3gacyb3ta]

Woah, sorta like IRC!

[elendee]

different level of abstraction... IRC is a protocol where devzat is an SSH application

[quackduck]

What server info should be included in a reply to such a ping?

[quackduck]

What do you mean by an owner handle? A GitHub/Twitter username?

[elendee]

I take handle to mean a "URI safe username", similar to a unique key for an asset in a URI often being called a "slug".

I don't think there's a hard definition. But I would call "bob" a username, "bob jones" a name, and "bob#1234" a handle

So yea an email could fall under that category. But since these will likely be public, I'd vote against that

[quackduck]

Yeah we can let the owner decide what they'd like to show. I think we can have this work over HTTP as JSON. We'd need to decide a default port for it.

[quackduck]

I wonder if we could detect an HTTP client on the SSH port and reply with this message instead of the usual SSH exchange.

[elendee]

That does seem ideal yea.
I think the next best option would be digging into some server-side language that has an SSH package, so you could expose it via that. That's what I did for MUD servers, using the node telnet package to repackage all the telnet data into a normal node server. I don't know if SSH has good packages for that though

[CodeLongAndProsper90]

You can access argv with an ssh command, correct? Just add an option called '--dump-info' and send the info as json across stdout