[movitz-devel] Driver problems

[Frode Vatvedt Fjeld]

Joel Smith <joels at mobyfoo.org> writes:

> That's great! I think it's just what I need. I assume you are going
> to add an operator for writing to ports as well?

Yes, that's the plan.

> I am a little worried about the lack of a defined API for the device
> code. I worry that when people start writing drivers for Muerte then
> they will all start doing their own thing in terms of the API. I
> realize that you wish to have only a "very thin abstraction over the
> hardware" and not a proper driver, which I tend to agree with, but
> don't we still need some sort of low level API for the various
> classes of devices.  If/when in the future someone comes to write
> proper drivers for a LispOS based on Movitz they are going to have a
> terrible time trying to support all the Muerte device code which
> shares no common interface. What do people think about this?

I think there's a slight chicken-and-egg problem here: It's hard to
specify an API before we have experience with writing a few drivers. I
think that the process will be like write a few drivers, tighten up
the API and update existing code, and so there is a decent API for
when/if the code-base grows. In my experience it's not so dificult to
change existing code.

So for example, I'm expecting to remove the io-read/write-sequence
functions, which now seem to me quite ill-designed.

-- 
Frode Vatvedt Fjeld