Re: Overlay mechanic improvements

[David Kastrup]

Richard Stallman <address@hidden> writes:

> [[[ To any NSA and FBI agents reading my email: please consider    ]]]
> [[[ whether defending the US Constitution against all enemies,     ]]]
> [[[ foreign or domestic, requires you to follow Snowden's example. ]]]
>
>     Because you don't want _anything_ that text properties do.  You don't
>     want to have the stuff cut&paste,
>
> Why not?  If the images are thought of as part of the buffer contents,

They most emphatically aren't.  Search and replace works on the
underlying text (invalidating the image when a replace happens),
incremental search works on the underlying text (temporarily removing
the image), saving saves the underlying text, all editing commands apply
to the underlying text.

> I'd expect it to be desirable that they follow text that is copied.
> If you kill text that contains some of these images of math
> and then yank  it back, shouldn't it come back with the images?

No.  The same input text at a different location in the same document
might show a different equation number.  In a different document it
might not compile at all or have a different size.

> This is what lead me to think of text properties first for this job.
>
>      you don't want to have the buffer
>     modified because images are switched on and off,
>
> We can provide a feature for turning off and on the display
> of these images without changing the text properties themselves.
> I think there already is a way.
>
>               you don't want anything inserted anywhere inheriting
>     anything from it.
>
> We already have ways to specify no inheritance.
>
>     you most definitely
>     don't want ever to split its identity in two if there are insertions in
>     the middle,
>
> That's true.  However, isn't such insertion anomalous anyway?  How bad
> is it, if that anomalous act causes the image to appear twice?

Bad.  Text properties and overlays have different semantics.  You
basically say "let's ignore all the mismatching semantics because the
performance of overlays is bad".

That does not make sense.  If it did, we would not have overlays in the
first place.

-- 
David Kastrup