Open Stack

On Tue, Nov 12, 2013 at 01:59:35PM +0000, Alex Heneveld wrote:
> This new proposal [1] looks like a very positive change on top of
> [2].  Simpler both to read and write.

Agreed, thanks Thomas for taking the time to write this up, it looks like a
good reflection of what was discussed last week to me.

> One minor suggestion is to consider using a special character (eg $)
> rather than reserved keywords.  As I understand it the keywords are
> only interpreted when they exactly match the value of a key in a
> map, so it is already unlikely to be problematic.  However I think
> it would be more familiar and clear if we instead used the rule that
> any item (key or value) which _starts_ with a $ is interpreted
> specially.  What those rules are is TBD but you could for instance
> write functions -- as either `$get_param('xxx')` or `$get_param:
> xxx` -- as well as allow accessing a parameter directly `$xxx `.

So, there has been some discussion on this topic related to the str_replace
usage of $ prefixes.  In the end we decided straight string replacement was
better in that case, see:

https://review.openstack.org/#/c/47772/

I'm -1 on any backwards incompatible HOT syntax changes at this point,
unless they can be very strongly justified - we did add a disclaimer to the
Havana release notes saying HOT was an unstable in-progress format, but
people will be using it nevertheless.

If possible, I'd prefer to focus on adding stuff to enable the
software-config functionality to work, in as backwards-compatible way as is
practical.

Steve