home *** CD-ROM | disk | FTP | other *** search
- >Maybe these keys could be caught by X and it could close kbd? dunno it
- >will need to know when to reopen them too I guess. I was connsidering
- >writing a CX like daemon when I started messing with ite.c, not there yet.
-
- Okay, I don't know how ite/grf/view/kbd/mouse are configured right now,
- but, I was just thinking about this problem and thought maybe I should
- toss some ideas out for arguments' sake.
-
- What if, in addition to /dev/viewXX, there were /dev/mouseXX and
- /dev/kbdXX devices (where XX are numbers). From an application
- standpoint, you open exclusive inputs and outputs (with matched
- numbers). The kernel must then provide a user focus for each of these
- devices. The focused view would be visible on screen, the focused
- keyboard would get keyboard information (make sure focus changes only
- happen when there are no keys pressed??), and the focused mouse would
- get mouse information (again, make sure no mouse buttons are pressed).
-
- It should be possible to change these focii in a configurable manner.
- Some people will want Motif like changes where keyboard, mouse and view
- change together, others will want Intuition like changes, where keyboard
- and mouse change together, independantly from view. I'm sure someone
- will want to be able to use their mouse on one view, type on the
- keyboard in another, and look at a third...
-
- Things that really need to be thought about: what about multiple
- displays? How about when the view focus changes, it affects only the
- display to which the view is associated, leaving whatever was last
- active on the other display.
-
- Actual focus changes. Perhaps a special keyboard device which can be
- programmed to trap only certain keys before they get sent through the
- /dev/kbdXX interface. Then, user configurable keystrokes could be
- watched for by a daemon to change keyboard/mouse/view focii.
-
- >There are no sprites :^( Sprites didn't figure quickly into the RTG
- >system I had dreamt up. I still have problems with RTG, let alone
- >trying to get a sprite into the system. :^)
-
- Well, since at least one graphics device (cc) supports a hardware
- sprite, I think support for them should be included. Since one graphics
- device that supports hardware sprites draws them using huge pixels
- (again, cc) it should be configurable. I think it is part of a view to
- know whether a hardware cursor can be rendered or not, and to handle
- that rendering. From an X server, it means if'ing out (or using
- pointers to functions or something) the cursor rendering code.
-
- Another thing to think about is a DUALPF screen - provide a monochrome
- overlay akin to certain sun graphics boards.
-
- >Use of the CPU is much faster than the blitter on all machines that
- >NetBSD currently (and probably hereforth) runs on.
-
- Yes, but, by the same token, I'm sure on a fast enough machine I could
- write a PIO SCSI driver that is at least as fast as DMA. The difference
- is PIO consumes many cycles from other things the machine could be
- doing, whereas DMA happens without much processor involvement. Same
- with the blitter. If I decide to turn OpaqueMove on in twm (once I get
- NetBSD and X installed and running...), I want the blitter to move the
- windows (if possible) not the CPU - that way everything else in the
- system can stay running.
-
- Just my thoughts.
-
- >Chris.
-
- Gregory Kritsch | 3A Computer Engineering at UWaterloo
- ggk@tirith.uucp | "Life is a highway
- ...!xenitec!tirith!ggk | Life is a stinking highway!" -BNL
-
-