chac@aratar.UUCP (Chuck Clanton) (10/26/88)
i cannot figure out how to implement a particular interface in X.
i have been using window systems for a number of years, and still find that
with most windowing interfaces i end up typing into the "wrong" window a
disconcerting amount of the time ("focus" errors). the feedback i currently
have available from X includes 1) change in title bar appearance, 2) change
in border appearance, and 3) change in cursor appearance (sometimes). when
i am intently concentrating on a task, these are not adequate to prevent
errors. there was one window system where i never made this unpleasant
mistake--rob pike's mux on the blit. since that was a click to focus
interface, i wondered if the positive click selection was the key. after
experimenting with this under X, it didnt seem so but perhaps it was some
detail of the blit's particular click to focus mechanics. recently i have
converted to using pike's sam editor, which has many similarities to mux
under X, and i still make focus errors. it must be something else.
another feature on the blit was that every window except the visible part
of the current window had a very light stipple--a dot on a 1/2 inch(?) grid
over the entire window except the visible part of the current window.
perhaps the state of the window i was looking at was always subliminally
apparent because of this. to find out if this was the key factor, i have
thought about trying to implement this under X. but i cannot figure out how
or where to do it.
my first thought was that this is a window manager issue. if there were
windows with appropriate transparency properties, i can imagine doing this
in much the same way that some window managers maintain title bars except
with an overlapping rather than the mostly obscured window of the title bar.
but no such windows exist. i dont know if a window manager can insert
itself between the application and the server enough to somehow impose its
will on every display operation, but that seems philosophically wrong and
likely to fail in subtle cases. of course, one could extend the X server
to do it i suppose...but isnt that a bit like saying "let's put it in the
kernel" which might be attractive in each individual case, but is extremely
unattractive in the sum of all individual cases.
can anyone suggest a way to do this? i am not particularly worried about
performance, i am just interested in testing the idea right now. or do you
have other suggestions about how to provide better feedback about the focus
window? replies that do not seem of general interest can be sent directly
to me, and if there is interest and enough material, i will summarize to
the net.jps@cat.cmu.edu (James Salsman) (11/01/88)
In article <317@aratar.UUCP> chac@aratar.UUCP (Chuck Clanton) writes: > i have been using window systems for a number of years, and still find that > with most windowing interfaces i end up typing into the "wrong" window a > disconcerting amount of the time ("focus" errors). I think that this is a fundamental stimulus-response compatibility issue. Click to focus (ala Macintosh) works well for beginners as well as those who are well-aquanted with the interface. Mouse position for focus trips up experts and novices alike. But, the best solution should be to select keyboard focus from the KEYBOARD, ala the TI Explorer. This lets experts WIN BIG in time-for-task, but you have to teach keyboard commands to novices who complain "why can't I use the mouse?" This isn't a problem on the TI since windows usually take the whole screen thereby completly covering the other windows. Keep in mind that the Explorer's programming interface is designed for programmers, and if you develop a product for a USER, the interface is up to you. :James -- :James P. Salsman (jps@CAT.CMU.EDU)