test for a (really) visible frame? test if a frame is the only (really visible) one?

[Drew Adams]

IIUC (please correct any misconceptions - consider these as questions):

1. Every terminal frame is considered visible, always, whether or not you can
see it currently.

2. `visible-frame-list' (and `frame-visible-p') can include not only frames that
you actually see but also (terminal) frames that you cannot see.

3. `emacs --daemon' adds an extra terminal frame named (usually? always?) `F1'
that is not (necessarily? ever?) visible to the eye, but is considered visible
by Emacs (since it is a terminal frame).

It seems that it is _never_ visible to the eye: If you delete (only) the other
visible frames, then you can no longer see any frames at all.

4. It is impossible to test whether a given frame is a terminal frame. There is
no predicate for this and no frame parameter that indicates this.

5. It is impossible to test whether a given frame is a/the hidden-to-the-eye
(but "visible") frame that is used only to implement `--daemon'.

6. It is impossible to test whether a given frame is the sole frame visible to
the eye. 

One could try to test #6 using (cdr (visible-frame-list)), but that will
apparently never be nil if `--daemon' is used, because of the extra,
under-the-covers frame, F1.

One could try to test whether, if `visible-frame-list' is a doubleton, the
second (or is it sometimes the first?) frame has the name "F1", to try to handle
the `--daemon' case. But that's fragile. Anyone could name any frame "F1" - the
name is not a reliable indication of the kind of frame it is.

[Caveat: I haven't tested any of the `--daemon' stuff myself, since I'm using
Windows, and AFAIK `--daemon' isn't supported on Windows. This is my
understanding based on hearing from others and reading the doc.]

---

Request: If the above are correct, could we please have some way to effect the
tests #4, #5, and #6?

Simple use case: Define a function to delete the selected window or, if only one
window, the selected frame - but never delete the last frame visible to the eye
(including when `--daemon' is used).

It seems like the following choices together have resulted in a fragile
situation that doesn't make it easy (or even possible?) to know what's going on:

a. Considering all terminal frames as "visible". (Why?)

b. Naming the terminal frames with mundane names like "F1". (Why?)

c. Not having any frame parameter or predicate that distinguishes a terminal
frame. (Why not?)

d. Adding an invisible "visible" frame, F1, as part of the way to implement
`--daemon'. (Why?)

I can understand (a) and (b), I guess. In some sense, terminal frames _are_
always visible. And it's useful to refer to them using simple names.

But things seem very different for the extra frame that is used to implement
`--daemon':

(a) It is not there to take over and be shown if you delete all other visible
frames. If you do that, you lose all frames visible to the eye - you lose any
way to interact with Emacs other than by the external command line.

(b) A user will never, I think, refer to that special `--daemon' frame by name,
switching to it as `F1' (is that even possible?). But unlike hidden buffers
(e.g. ` *foo*'), this frame's name is a normal one.