Re: Design of instructions for Counter/Timer subdevices...

On Fri, 11 Apr 2003, Frank Mori Hess wrote:

[...]
> > Anyway, I think the double use of the instruction function leads to
> > too many misunderstandings. Already now, configuration instructions
> > misuse the "data" fields, because their type is lsampl_t. Not really a
> > good situation...
> >
> > Giving counters their own data structure requires a change of the
> > Comedi API, but it would be a good occasion to get rid of the
> > "chanspec" integer; in fact, this integer is already a mini
> > data-structure forced into 32 bits. Giving each channel a real
> > configuration data structure would make things much more
> > understandable, I think. At the same time, a new configuration
> > function should be introduced: comedi_channel_config(), for example.
> 
> The real issue here is that comedi has achieved a state of eternal beta.  
> What we really need is to reach 1.0 and then we can branch off the stable 
> version and start making tweaks to improve the API.  What might help is to 
> clearly define exactly what needs to be done for 1.0.  Maybe through using 
> the bugzilla, and assigning various bugs to comedi/comedlib version 1.0? 
> 
I don't see the added value of moving the discussion away to the
bugzilla... Moreover, what I said has nothing to do with "bugs", it's
an API and design issue.

I'm not pretending at all that my suggestions are good. They are
nothing more than bringing David's own suggestions back into
the spotlights, trying to put a step forward.

Herman

-- 
  K.U.Leuven, Mechanical Engineering, Robotics Research Group
<http://people.mech.kuleuven.ac.be/~bruyninc> Tel: +32 16 322480

Received on 2003-04-11Z19:35:25