SSH access over a slow network connection?

Why not proxy the network calls only then and use a real browser?

Use SSH’s dynamic proxy mode (-D) so you have the full browser running locally and only the network traffic running over SSH.

This is enormously better for UI response time and things like compressed files: you only transfer the compressed text, image, audio, or video files rather than the decoded form which can be an order of magnitude larger.

Character based interfaces are just nice, IMHO. The Borland IDE was the pinnacle of dev interfaces IMHO. It updated to have modern language server support etc would be my dream dev environment. It would be cool to see how far a compliant web browser could be taken in text mode. It's really sad that most terminals don't have proper image support, btw.

Edit: A comment here linked to https://www.brow.sh/ It looks amazing.

  > but why would you render to a low-resolution 2D buffer with random character-hacks

Have you used a terminal in the last 5 years? We got graphics and everything now[0]

  > huge amount of overhead

A... terminal... with... large overhead? Surely you're joking. Have you met any gui app? Microsoft word takes a few gigs. Have you seen electron apps like Spotify or slack? My fucking mail client takes over half a gig and neomutt takes almost nothing.

Either you're making numbers up or we live in different realities

  > actual hardware acceleration?

Hardware acceleration has existed for years...

Maybe stop making assumptions and check out what terminals can do. They've come a long way since the 80's. I'd suggest starting with something like ghostty.

[0] https://notcurses.com/

Wouldn't you still be as sluggish as the GUIs you mention since it's HTML5? Is the benefit from only rendering a subset or something? I'm struggling understanding how rendering a page into a terminal is faster than just rendering to a regular window. It just seems like another layer between the browser and the GPU.

I think you have a misunderstanding of the full system. Like you are thinking the GUI is much simpler than it actually is. Conversely I think you're thinking that CLIs are much more complicated. There's a lot more going on than just painting pixels on the screen. Maybe start by asking why vim is several magnitudes less resource intensive then Microsoft word. Then ask the same about neomutt vs Thunderbird vs outlook/Apple Mail

I'm also not sure why you think it's an extra layer between the gpu. Both the browser and terminal emulator are sitting in top of the OS. But also remember if you run headless then your OS doesn't need to load the desktop environment, though you can still use the CLI. At this point clearly the GPU is seeing less, right?

I understand that a GUI is complex, which is why I'm confused about the performance benefit as HTML5 already acts like a GUI. Even if it's a terminal program, it would still need to calculate the layout and handle the DOM to remain compliant, no? A website can be as intensive as Word.

My logic with it being an additional layer is that instead of the browser rendering text to the screen, it would have to speak with the terminal for the terminal to instead render text to the screen. Same goes for rendering graphics. Besides a headless system I don't see much of a benefit to it.

Just less overall overhead

  > A website can be as intensive as Word.

Word has no business being as intensive as it is. It is a bloated piece of garbage.