Re: Talk driven experimentation

> The first change is that I tried adding a code field to a <host>, for
> example <reboot>, with the expectation that doing something like:
> 
>     % ark host reboot slimy
> 
> would reboot 'slimy'. Unfortunately, this isn't what happens. The engine
> gets it's pants in a twist over the host context and the proxy host. ...

Smells like a bug.

This kind of thing is supposed to be controlled by the
once-per= and proxy-hosts= attributes of a <code>.

once-per= specifies how many "places" (specified as a table
of proto-hosts) that the method must be run for it to be
fulfilled for the whole site; proxy-hosts= specifies the
(real) hosts that are prepared to act for the proto-hosts.

Maybe we need a small refinement of this spec?

Could you check in your code, and also give a quick
checking-for-lies spin through "Method granularity
(once-per) and proxy hosts" in
http://ark.sourceforge.net/ark-conflang.html (source:
ARK/webpages/ark-conflang.htma), on the dubious theory that,
if we make the documentation clear, the code might follow
:-)?

> The second thing I've done is remove a couple of error checks out of
> ArkControl (control.py). ...

Pls check it in; we can deal w/ any consequences...

> (PS: ARK runs fine on Mac OS X, so I can work on the sofa
> in PowerPoint editing slides, edit XML files in BBEdit,
> then switch to the Terminal and fire off an ark command
> which wirelessly ssh-es into my other portable sitting
> upstairs and instructs it to do something. Very groovy -
> expect to see this magic at the talk.)

(I worry about that boy, I really do...)

Will