README.md 5.77 KB
Newer Older
oneM2Mdev's avatar
oneM2Mdev committed
1
2
3
oneM2M-schemas
==============

PeterNiblett's avatar
PeterNiblett committed
4
5
6
This is a private repository for developing oneM2M xsd files

The default branch will initially contain a "0.80" version of the  files, mathching the 0.8 version of TS-0004
PeterNiblett's avatar
PeterNiblett committed
7
8
9
10

Further work / questions on the schemas is listed here. Many of these need CRs against TS-0004

General
Peter Niblett's avatar
Peter Niblett committed
11
12
13
 
    How should virtual resources be represented? According to TS-0001 they are Child Resources, but they have a well-defined name which makes them look a bit like attributes that have a URI type. Presumably we are supposed to return these URIs if rc includes "child-resource-references" but not if it asks for just "attributes" or for "child-resources". So one interpretation is that we don't show them in the schemas at all, and they are just returned via the childResourceRef mechanism (this would mean we would need to add values for the virtual resources to the resourceType enumeration). The alternative is to represent them as explicit named elements in the schema (like we do for attributes) and return their values that way (when rc permits this). 
    There are a few places where attributes are defined as xs:string. If these are human readable - e.g. groupName - that's fine (though it raises a question of whether they should carry the xml:lang attribute). In some cases - e.g. contentType or locationStatus - it seems that the value is meant to be machine readable, but we don't enumerate (or even define) the allowable values for them.
Peter Niblett's avatar
Peter Niblett committed
14
    Showing the cardinality of child resources is tricky for resource types where the cardinality varies by child resource. It's easy to do if we make them a sequence, but that forces a particular order. If we want to allow them to appear in any order, we have to use a repeated choice, which makes it hard to force a minOccurs of 1 or a maxOccurs of 1 on any of the of the child resources - they all become [0..n] 
PeterNiblett's avatar
PeterNiblett committed
15
16
17
18
19
20
21

\<accessControlPolicy\>

    Do we need to give explicit names to all the types currently defined in the file?  
    Should some or all of theses types go into CDT-CommonTypes or CDT-Enumerations?
    Since accessControlOperation is an enumerated integer type, we could represent accessControlOperationList as a   list of simple types. This would be shorter than having it as a complex type.
    Look at definition of accessControlOriginators
Peter Niblett's avatar
Peter Niblett committed
22
    Do something with the 2 character country codes.
PeterNiblett's avatar
PeterNiblett committed
23
24
25
26

\<container\>

    What values go in latest and oldest if there are no child resources?
27

Peter Niblett's avatar
Peter Niblett committed
28
29
30
31
\<CSEBase\>

    m2m:resourceTypeList can be removed, since this type can be declared inline

Peter Niblett's avatar
Peter Niblett committed
32
33
34
35
36
37
38
\<execInstance\>

    It isn't clear what the type of execDisable ought to be, as it isn't obvious what the cancellation mechanism is. TS-0001 says "The UPDATE request shall address the execDisable attribute with a predefined value in order to trigger the CANCEL action" i.e. you trigger a cancellation by writing a "predefined value" to this attribute, but doesn't say what that value is. We could achieve that by making it xs:boolean or by making it a string with two permitted values (e.g. "" and "CANCEL"). On the other hand TS-0004 says "Cancel operation is triggered by an Update primitive, if the primitive addresses the execDisable attribute or the URI provided as the value of the execDisable". This suggests that execDisable should be xs:anyURI and that triggering is cancelled either by writing a new URI into it, or by sending an Update to the URI that it already contains. The corresponding execEnable attribute in \<mgmtCmd\> has type xs:anyURI.

    In TS-0004 7.2.16 the execStatus attribute has type m2m:execStatusType, but in the enums clause the type is called m2m:execStateType


Peter Niblett's avatar
Peter Niblett committed
39
40
41
42
43
\<group\>

    The order in which the attributes appear is different between TS-0001 and TS-0004 (also TS-0004 should say creator not Creator) 
    Should the element for the virtual resource be called fanOut or fanOutURI?

Peter Niblett's avatar
Peter Niblett committed
44
45
46
47
48
\<node\>

    There's an attribute which is called called hostedCSEID in TS-0004, but is called hostedCSELink in TS-0001
    The cardinality of the child resources is enforced (see general comment)

49
50
51
52
\<serviceSubscribedNode\>

    Needs to be added to TS-0004 (in place of authororizedNode)

Peter Niblett's avatar
Peter Niblett committed
53

PeterNiblett's avatar
PeterNiblett committed
54
55
\<subscription\>  

Peter Niblett's avatar
Peter Niblett committed
56
    Should the eventNotificationCriteria type be moved in to Common Types? (i.e. is it used anywhere else)?
PeterNiblett's avatar
PeterNiblett committed
57
58
59
60
    
CDT-enumerationTypes

    Add or update definitions of the following types:
Peter Niblett's avatar
Peter Niblett committed
61
62
      resultContent, statusCode, requestStatus, notificationCongestionPolicy, SRole-ID
    Decide how to represent the attribute attribute
PeterNiblett's avatar
PeterNiblett committed
63
64
    Does m2m:memberType need to be different from m2m:resourceType? Can we remove it?
    Sort out the two m2m:operation types
Peter Niblett's avatar
Peter Niblett committed
65
66
    Should have a better name for listOfBoolean (it isn't a list)
    The third value of m2m:consistencyStrategy is called MODIFY_TYPE in TS-0004, but SET_MIXED in TS-0001
PeterNiblett's avatar
PeterNiblett committed
67
68
69
    
CDT-commonTypes

Peter Niblett's avatar
Peter Niblett committed
70
71
72
73
74
75
76
    There are two different types to represent a list of URIs.  One is a simple type (using XSD list) the other one is a complex type, containing a separate child element for each URI. 
    We should look at the way that IDs are represented in the schemas. At present most of them are just xs:anyURI. Is that sufficient?
    There is some confusion as to whether the type is m2m:id or m2m:ID (and whether the other xxxid types should be xxxID or not).
    TS-0004 now seems to have three different kinds of timestamp.  There's m2m:timestamp, xs:dateTime and xs:dateTimeStamp
    It is not easy to see how to represent the changes made by PRO-2014-0562-Implementation_of_Filter_Criteria_as_concept
    Consider moving the m2m:filterUsage enumeration type into CDT-enumerationTypes
    The order of elements in m2m:operationResult looks a bit strange
77
    The declaration of m2m:responseStatus/description looks wrong
Peter Niblett's avatar
Peter Niblett committed
78
79
80


    More general revision and checking of CDT-commonTypes is required (defer this until after the other XSD files are complete)
PeterNiblett's avatar
PeterNiblett committed
81
82
83
84