Re: Getting $(allclasses) into a shellcommand

[Ted Zlatanov]

On Tue, 20 Jul 2004, ebrown@lanl.gov wrote:

> 2. When do too-long $(allclasses) strings get created?  So far, I've
> only seen it mentioned with respect to 'packages:'.  Defining classes is
> the only action available in 'packages:'.  It would be great if you
> could execute arbitrary commands, as in 'processes:'.  This could help
> eliminate some of the need to rely on class definitions.  This
> possibility was discussed in a thread several months ago,
> http://lists.gnu.org/archive/html/help-cfengine/2003-11/msg00061.html
> , and I wonder if the implementation might still be in the works?? 

I tried to shift the discussion to "how does cfengine pass data to
child processes."  Right now this happens through the command line,
which has many limitations compared to a regular data exchange
protocol such as the one I proposed.

Specifically for $(allclasses) this is not necessary, but maybe next
something from cfenvd will need to be passed to a child process, or
better knowledge of cfengine's internals (e.g. to tell the process if
it's started from shellcommands: or processes:) will be needed.  How
do you distinguish between all the data cfengine could pass down?
Instead of doing this ad hoc, it's better to set the standard now, I
think.  Of course it's not my decision, I'm just offering some thoughts.

Ted