I wish I had more evidence-based arguments for my opinion, but I strongly believe text-based interfaces are among the best out there. They have an ability to squeeze in an uncanny amount of information in any given space, without turning cluttered.
If I think about it, even GUIs tend to be driven by text in most cases. The difference is that every word gets a graphical border (be it in the shape of a button, a tab, a menu item, or whatnot.) These borders are needed in a GUI where the location of the text is less predictable, and it helps novices to see the window under a menu while the menu is shown, but these are not problems for experienced users at predictable text-based interfaces.
One really great things about text interfaces is they're (drumroll) text-searchable! That's one huge benefit of emacs, if you're shown a huge list of options, hop into the buffer and just search. I've been frustrated by some parts of microsoft interfaces where I'm presented with lots of text options but couldn't search them (can't remember where).
A note on the attached PDF document, there seems to be a total blind spot about readability being damaged by dumb colours, look at Figure 8 on P11, I can read the coloured text but it is noticeably harder, therefore slower, than reading the white text on a black background. Also the light blue on white for listing 13 as another example.
I think Emacs is an illuminating example, since its 'GUI' interfaces really outshine emacs being run in a terminal emulator. Text interfaces are great, but terminal emulators in particular are very limiting. Inlining images always requires hacks and even with hacks a terminal emulator can't do nice things like have multiple font sizes in the same frame (discounting double height/width modes, since they're so limited I've never seen them used effectively.)
It's not the case for me. I far prefer emacs in a terminal over the gui, and thats how I've used it for years. I can't run the gui on a gnu screen session on that server I've been working on, for example. I admit I'm probably the outlier on this though.
Nothing wrong with being an outlier but if you can explain, I'd like to know why you prefer it (when you can, that is).
I'm not mad on going mad with colours and effects, something that happens too much, but even some minimal visual feedback can be disproportionately helpful I feel.
I'm the kind of person who lives in the cli, so I love tui's, I use a tiling window manager and gnu screen heavily. For me, it just fits my workflow.
I have slowly been transitioning away from using emacs for everything though. I use it as my text editor but I no longer use it for rss, irc, email, etc. I would probably change my tune if I went back to that.
That is a very interesting point, and I'm with you on the extra facilities (typefaces, colours, highlighting, sizes, effects etc) that are possible and useful. I'll chew this over tonight. It's a good point.
Unfortunately, many things can break text search in web applications: components that render into DOM only when they are visible and in the viewport, components that are covered by something else so you can't see them (even though you can search them), pictograms etc.
Anything where information and/or buttons need to be presented to the user, really. Text-based interfaces tend to support a high density of either without loss of overview.
Part of this might be, as I allude to in my original comment, the lack of other visual structures that otherwise might distract from the content. Another part might be that humans are pretty darn good at interpreting text.
If I think about it, even GUIs tend to be driven by text in most cases. The difference is that every word gets a graphical border (be it in the shape of a button, a tab, a menu item, or whatnot.) These borders are needed in a GUI where the location of the text is less predictable, and it helps novices to see the window under a menu while the menu is shown, but these are not problems for experienced users at predictable text-based interfaces.