Extending the use of XTGETTCAP for feature discovery

[aymanbagabas]

Hi everyone,

The current state of terminal feature discovery is wonderfully varied! Each terminal has its own approach. Some suggest to use DA1, others use extended Terminfo capabilities, and the rest leave it up to the application to figure it out. I think that's a good opportunity to find a better solution.

I think Terminfo databases feel like a relic of the past, and ideally we'd move beyond them. But that's not quite how technology evolves 😄 Terminals and escape sequences grew over decades and still going strong, so backwards compatibility is important. Xterm, the modern terminal standard, defines XTGETTCAP for applications to query Terminfo/Termcap capabilities. While it was introduced originally to query "for special keyboard keys", today many terminals (including Xterm itself) use it to expose their Terminfo database capabilities. This is great because it allows applications to discover capabilities over the wire instead of bundling them or relying on Ncurses or other TI database packages to exist on the system.

That brings me to how it's being used nowadays. Several applications already use XTGETTCAP to query capabilities like RGB, setrgbf, and setrgbb. Other known user-defined capabilities that are also commonly used include:

• True color support Tc originated from Tmux
• Clipboard selection Ms (OSC52)
• Underline styles Smulx (SGR 4:N)
• Underline colors Setulc (SGR 58;2;...)

Tmux also defines these:

• Hyperlinks Hls (OSC8)
• Working directory Swd (OSC7)
• Progress bar Spb (OSC9;4)
• and others

Someone might ask why not DA1? The way DA1 maps each attribute to a number is quite limiting, and DA1 attributes also conflict with well-known OSC identifiers like OSC8. String based capabilities are much richer and a natural fit for today's world, where transferring few extra bytes is basically free compared to the days of the DEC VT100.

So I'd love to propose extending the use of XTGETTCAP to query terminal features. In particular the many OSC ones that cannot be queried via DA1/2/3, DECRQS, DECRQSS, DSR, or others mechanisms. I'd like to suggest adding Hls, Swd, Spb, and other commonly missing capabilities to Ghostty's Terminfo database so they become discoverable via XTGETTCAP. I think this would be a great step towards unifying the efforts to discover terminal capabilities by leveraging an existing, widely adopted protocol and well-established capability names.

[rockorager]

I did a search and no other common terminal exposes these as terminfo capabilities, except for tmux. That said, there doesn't seem to be a collision. I'm for this.

[j4james]

Just FYI, XTGETTCAP is not supposed to be reporting general termcap capabilities - only the keyboard keys and three special case values: Co (or colors), TN (or name), and RGB. This is an extract from the Xterm documentation (emphasis mine):

The string following the "q" is a list of names ... which correspond to termcap or terminfo capability names for special keyboard keys. A terminal description will include other capabilities... which are intentionally not part of this interface.

I realise Ghostty is already diverging from these requirements (and I think also not responding in the correct format), so if that's intentional then it doesn't matter what further changes you make to the protocol. However, if you intend to be Xterm compatible, and actually follow the XTGETTCAP specification, then there are probably some issues that need to be addressed. Adding additional capabilities to the report would not be helping.