[Date Prev] | [Thread Prev] | [Thread Next] | [Date Next] -- [Date Index] | [Thread Index] | [List Home]
Subject: name and nameID attributes on choreography/flow constructs
Hima pointed out to me that except for the Transition
element, each control construct has a name attribute group in its type. So Fork, Join, Decision as well as Start, Success and Stop
all have to have name and nameID values. I agree with Hima that we probably need to make a uniform
stylistic decision. I am not aware that there is any good use case anymore for
needing to refer to what used to be called the “pseudostates”. The well-formedness constraints on what types of elements
IDREFs can refer to generally prevent referring to these states. A no-op process (with no real states) could be formed by
letting a Start state reference Success termination. That is not a very useful
use case, however. I am willing to leave names and nameIDs in if people think
they might be useful. If we do that, I would prefer to add the name attribute
group to Transition, just to make it easier to always do the same thing. I am also willing to remove the name attribute group from
all these states (which leaves me less comfortable somehow). For fewest changes, we can add the attribute group to
Transition, even though these things are less than first-class entities. We can either remove them all during our TC review or when
we update the version to 2.1 or 3.0. Finally, we could introduce a second attribute name group,
just like our current one, except making the attributes optional. That might be
the best way to allow their use if and when needed but not to clutter the
notation with a lot of attributes that are infrequently used. Actually I like that last solution the best. I would like to
see some +1s before doing this as it seems slightly more than mere editorial. Dale Moberg |
[Date Prev] | [Thread Prev] | [Thread Next] | [Date Next] -- [Date Index] | [Thread Index] | [List Home]