Re: Gnopernicus Sending Output to Serial Device Crashes Hardware Synthesizer

[Bill Haneman <Bill Haneman Sun COM>]

Peter Korn wrote:

> Hi Bill,
>
> I'm unconvinced of the importance of having Gnopernicus talk to 
> Braille devices by default - at least to talk to Braille devices 
> *directly* by default (vs. via BrlAPI). 

I agree that using brlapi for this makes the most sense, as the 
autosensing/probing support can be offloaded to a package which 
specializes in braille hardware support.

However I believe that gnopernicus already does some auto-sensing of 
braille displays.  If not, then I think that the default configuration 
should be changed so as not to send braille to ttys0 without having been 
explicitly configured to do so.

The most serious issue seems to be that the DECTalk Express can be 
gotten in to a persisting bad state (i.e. one that doesn't respond to 
mere power-cycling) with a certain serial stream.  Since it is not 
reasonable to expect one device driver (e.g. braille) to understand the 
behavior of an unrelated device (e.g. synth), I don't see a general 
solution to this problem.  If a device is capable of getting put into a 
bad state via serial input, there is no way to guarantee that no other 
driver will ever emit that input 'coincidentally', even as part of an 
exploratory device 'probe'.

- Bill