Danek Duvall wrote:
On Fri, Jul 10, 2009 at 11:54:17AM -0500, Shawn Walker wrote:

I'm fine with whatever the consensus is.  My feeling though was that it
was more user-friendly to accept policy overrides on the command-line
regardless of subcommand without having to maintain a list per-subcommand
on which policies were allowed, etc. and just having to error out to the
user.

I don't see what that has to do with making the option global vs
per-subcommand, though.  Just because the option is recognized in one place
or another doesn't mean it has to have any other different behavior.

Instead, the approach of "you may override or specify any policy values
you like, and if applicable, they will be used for the operation" seems
nicer.

And that can work even if you put the option flag after the subcommand.

Am I missing something?

No, but I was. After mulling over the subcommands again, I agree that there are a limited number of subcommands it would ever be applicable to.

In particular, my prior justification was that it would allow us not to
error out when --policy isn't available for a particular subcommand, and we wouldn't have to document on a per-subcommand basis the --policy option. However, with the right wording and arrangement I think we can achieve a happy compromise.

I also found one case where having --policy be at the subcommand level made more sense: iamge-create. In that case, --policy will be used to specify the initial values for the new image instead of acting as a temporary override for the operation.

It also doesn't make sense to use --policy with any of the property or policy subcommands.

The following subcommands are the ones I know have immediate applicability for --policy:

- install
- uninstall
- image-update
- fix

Eventually, there will be policies that will subtly influence these subcommands:

- verify
- list
- info
- contents
- history

I'll update the proposal accordingly.  Thanks for sticking with this.

Cheers,
--
Shawn Walker
_______________________________________________
pkg-discuss mailing list
[email protected]
http://mail.opensolaris.org/mailman/listinfo/pkg-discuss

Reply via email to