Re: [obix-xml] Schema update

[Aaron Hansen <ahansen@tridium.com>]

Lets talk some more about batching at the next telecon before you go for 
it on the multiple poll request/response.

I think all top level messages should return an error (even 
unsubscribe), and thus, locale would necessary for each.

We'll talk about the other issues in Dave's email at the next telecon.

Some additional schema notes:
1. Error, include, unit, object, and range item types all have id and 
display elements.  What I'd like is for id and display to always have 
the same sequence location, no matter what the parent element.
2. Default values are needed for boolean elements of the include type.

Regarding the code generation difference between the 'has more children' 
options.  Since there really isn't any significant difference and only 
one option is schema enforceable, the solution seems obvious to me: 
<children><more>true</more></children>

Aaron


drichards@trane.com wrote:

>I posted an updated schema.  Apparently I could have done it as a revision to the original, but I didn't know that so it's a separate doc.  Here are some comments to go along with it:
>
>This schema includes cleanups and corrections from the .5 version.  The only
>known deviation from the documentation is the <ext> element of the read result.
>The base messagetype is defined but currently NOT used.  There have
>been some new comments around 'locale' which is the only element currently in
>the base.  Also, of the twelve messages defined, three do not use 'locale' and
>three others already inherit existing message types.  So it seems a bit
>premature to say there is such thing as a generic messsage type.  Is the
>additional complexity warranted right now?
>
>From the last meeting, what do we want the multiple Poll request to look like - or more
>interestingly, the multiple Poll response?  Should I just do it and then we can discuss the result?  Does this mean there is also a multiple 'Unsubscribe'?
>
>Should 'Unsubscribe' be able to return an error?
>
>What is the philosophy for lists/arrays?  In our current scheme, each element must be a different id.  Do we want an indexing method?
>
>Do we really need the first level <object> element in the write request?  We could just use the <children> element.
>
>Another interesting note:  For the min and max facets, the values are many times not fixed but rather related to other points.  Do we want to have a 'ref' attribute or something to say that the value is dependent on another point?  Still providing the current value, of course.  The issue will be that if you also try to change the other value, the indicated min/max values may no
>longer be valid when you try to write.
>
>
>Items that may want to be more restricted than xs:string:
>
>1) All the 'id' elements and atributes
>2) ParamType/Name attribute
>3) UnitType/Symbol
>4) enum element used for values
>5) ServiceType/Service element
>6) subscriptionId elements (all uses)
>7) vendorUrl and productUrl elements of About Response
>  
>