[coyotos-dev] More progress, some design issues

[Jonathan S. Shapiro]

On Mon, 2007-05-07 at 02:01 +0200, Valerio Bellizzomi wrote:

> So you are inclined to in-process fault handler, or lesser UPCB. It sounds
> like a good trade-off ...

Well, I'm currently coding down that path. This is a case where the
right decision is best made by trying the implementation. What happened
here is that we abandoned the UPCB for one reason (difficult of register
management) and discovered during implementation that we had created a
bunch of other issues by doing so. I am now trying a middle position:
take the register set out of the UPCB but leave the invocation and
activation frames *in* the UPCB.

Perhaps there will turn out to be a decaying body buried down this path
as well. I hope not. :-) It didn't seem like a good use of energy to
revert the spec until we know if the implementation is likely to work.

> There are two missing images in the spec, at "6.4.2 Meaning of String
> Items (stringitem_t)".

Yes. sorry. The input document had the wrong source file type for the
image. I have fixed it. THANKS for letting me know about this!


shap
-- 
Jonathan S. Shapiro, Ph.D.
Managing Director
The EROS Group, LLC