On 20 July 2012 15:53, Ben Pfaff <b...@nicira.com> wrote:
...
> For the string representations, as I told Joe Stringer, I'm leaning
> toward requiring them to be in the correct order for instructions listed
> above. Otherwise it seems risky that one could confuse the user.
> Similarly for the OF1.0 extension representation of OF1.1 instructions.
To this end, I've just been considering the way that instructions are
parsed. My understanding is that internally, we work towards having no
distinction between actions and instructions. The internal format
should be correct (instructions follow actions), if we are parsing
inputs correctly. As I understand, this means there is no motivation
to have a delimiter as in patch #2.
So when we parse, we enforce the ordering -- I would expect in
str_to_ofpacts(), the logic would go more or less as follows:-
while (ofputil_parse_key_value(...)) {
/* if current key is an instruction, break; */
/* parse as an action/same logic as currently. */
}
foreach (instruction_type t) {
/* if current key is not an instruction, error() */
if (current_key == t) {
/* parse */
} else if (key < type) { /* i.e., if the user specifies an
instruction in the wrong order. */
/* bad ordering or duplicate instruction: error() */
}
ofputil_parse_key_value(...)
}
if (left) error();
The above logic would force the user to put instructions/actions in
the correct ordering, rather than the solution of patch #6 which
allows any ordering, but may be executed differently than the user
expects. Is this more in line with the direction you're expecting,
Ben?
Thoughts? Comments?
_______________________________________________
dev mailing list
dev@openvswitch.org
http://openvswitch.org/mailman/listinfo/dev
