Re: [MiNT] XaAES / GEM memory issues

[Sven Karlsson <svenka@it.kth.se>]

Hi,

"Konrad M. Kokoszkiewicz" wrote:
> This only means that the top application, which received the event, must
> reroute the event to proper place after discovering that there is none of
> its windows at the place where the user for example clicked. Still this
> can be done by the event library (running in user context) and there is no
> need to put all this into the device driver.

1) read my post about how many events you can get from the mouse
2) Any AES application can get all mouse events. Not only mouse clicks
but events for any mouse movements. These events are sent possibly as
soon as they arrive from the IKBD.
3) Passing events from applications to applications increase the number
of context switches. In the worst case multiplying it with the number of
AES applications.
4) This would make the AES much more fragile than it is to day. Each AES
application must know of the the other AES applications and chose which
to send the events to. What if the top application has crashed? The
_entire_ window system will _halt_ immediately. Not to mention how hard
it would be for the AES library to know where to send the events.

Sorry, I don't buy this.

I stick by the design in the preliminary Fenix AES specification. IMHO
It has to be that way and in MiNT it must be implemented as a device
driver or kernel module for performance reasons. But then again if
someone propose something better I'm all ears.

Regards
 Sven