Re: ocp's/stations and their properties [message #325 is a reply to message #321] |
Mon, 02 July 2012 07:05 |
Christian Rahmig
Messages: 151 Registered: January 2011
|
Senior Member |
|
|
Hello Dirk and anyone interested,
> For my understanding of the concept of <ocpGroups>, please check whether
> the following statements are right:
>
> - An <ocpGroup> cannot be referred to - all other RailML elements refer
> to <ocps> but never to <ocpGroups>.
Well, for my understanding it should be possible to refer to an
<ocpGroup> as well, because it has the same attributes like an <ocp>
element.
> - An <ocp> can belong to more than one <ocpGroup> (which means: can be
> referred by more than one <ocpGroup>).
This is basically possible, but is that a realistic case?
> - An <ocpGroup> can have all the attributes of an <ocp>. There is no
> attribute of an <ocp> which wouldn’t also be available as an attribute
> of an <ocpGroup>.
Yes.
> - An attribute defined at an <ocp> can be overwritten by a corresponding
> <ocpGroup> (which means: by an <ocpGroup> which refers to that <ocp>).
I think the concept of overloading attribute values needs to be turned
upside-down: The attributes defined in the <ocpGroup> should be
overwritten by corresponding attributes from a referenced <ocp>.
> - An attribute defined at an <ocpGroup> is valid for all its <ocps>.
Considering the above concept, this will be only correct if there is not
a corresponding attribute in the <ocp> element itself.
> [...]
> I would prefer the other way ‘round: An <ocp> refers to it's <ocpGroup>
> (if there is one). If necessary, an <ocpGroup> can also name it's <ocps>
> but this would mean redundancy (crossing links).
That is an interesting idea. If this bottom-up-referencing is better
applicable for the <ocp> modeling, we should adapt our first idea.
> Another question is: With <ocps> and <ocpGroups> having the same
> attributes: Why don't we allow an <ocp> to refer to another <ocp>? An
> <ocp> can act as an ‘direct’ ocp or as an <ocpGroup> (or both).
The idea of <ocpGroup> was supposed to follow the pattern of <track> and
<trackGroup>. However, your question brings it to the point: Do we need
an explicit <ocpGroup> or may we reference from one <ocp> to the
next/parent <ocp>? What do others think about this question regarding
their usage of <ocp>?
> So, my suggestion would be:
> - to define an optional attribute ‘parentOcpRef’ (=tGenericRef) of an
> <ocp>.
Thank you, Dirk, for the approach.
Best regards
---
Christian Rahmig
railML.infrastructure coordinator
|
|
|