Need to provide formal description of behavior for operations. Inter-operability will be a far stretch without it (Larry Johnson, 20091204)

The lattice of the UML model is converted into a tree, all associations being converted to contained elements causing any shared elements to be copied into others.

Need to add revised text that documents the algorithm and regenerate the XSD.
The single XSD becomes multiple due to the refactoring that resulted in the resolution of other issues

We currently attach RecordKeeper to ManagedRecord, but this was established before we discovered the need to manage record disposition through RecordSet. Is there a case wherein records in a RecordSet have different RecordKeeper's?

Delete association between ManagedRecord and RecordKeeper. Establish association between RecordSet and RecordKeeper.
Provide constraint that RecordSet's cannot be merged in less they have identical RecordKeeper history.
Added constraint to RecordSet: “RecordSets cannot be merged unless they have the same RecordKeeper”. This is probably also a constraint on the DispositionInstructions service.
There is an undesirable direction of dependencies. Split the Role subtypes out of the Party package into an RmsRoles package that extends the Party package.

A document can be shared among multiple ManagedRecord's. The sharing is predicated on the immutability of a Document. CaseFilePart's are permitted to be appendable, deleteable, etc... all modifications of the Document. If a Document is shared among these MR's, constraints must be put on the characteristics of the CaseFilePart to assure to assure the integrity of the Document in the context of the vanilla ManagedRecord.

In examining this seemingly "innocuous" issue, it became clear that our model was somewhat "upside-down". To resolve it, the CaseFile package will be eliminated and the concepts merged into ManagedRecord. Essentially a ManagedRecord is a CaseFile with immutable parts.

Need to provide formal description of behavior for operations. Inter-operability will be a far stretch without it (Larry Johnson, 20091204)

Out of Scope
Revised Text:
None

Currently records sharing a Category are placed into a common RecordSet awaiting a "Cutoff" operation based on an event which is often date related.

CaseFileRecord's do not generally "Close" at the same time or in the same period.

Add constraint that the CaseFileRecord's that are in a RecordSet must be closed for that RecordSet to be Cutoff.

Are we missing a Repository Service? How do we handle the path were the record will be managed? (Bill Manago, 20091112)

We don't want to build in the idea of location and repository structure any more than is necessary. Specifying locations would defeat many of the goals of Cloud Computing, or even basic distributed data principles.
An implementation may gather information on a desired location (whether we recommend doing so or not) by passing it along as a value of an attribute defined in a AttributeProfile
Revised Text: None

RMS would be better cast as a MOF model as opposed to a UML model. Using constructs such as Association Classes cause difficulties in export to XMI and transformation to XSD's. (Raised by Sridhar Iyengar of IBM in the Jacksonville meeting)

We need to have the model remain as a UML model. This is not a metamodel on which models will be based, but is a model itself. The PSM will need to address the issue of Association Class representation in an XSD. That is being worked under Issue 15212
Revised Text: None

CaseFile is referenced in the static structure of the model; however, there are no services defined to capture or modify case files. Has this been further defined or should we assume we develop our own services and interpretation?

The requisite services are to be added

Do we need artificial root element to reach everything in the RMS.XSD

A similar construct was mentioned and used in the Federal Transition Framework XSD. It is not clear for example, that one could transfer ManagedRecords as well as Category for the ManagedRecord to refer to (we may only be able to transfer the Category reference, but not the description of the Category itself

Disposition Note: Merged into 50512
Resolution:
None
Revised Text: Unspecified

We can eliminate this by splitting out the party model (the only reference from ap to rms) into its own namespace

Split out Party and Document from RMSDomain package to eliminate the circularity. Rename packages RmsDomain to RmsCore, Party to RmsParty and Document to RmsDocument.

A number of errors are in the RMS.XSD. Some are namespace problems (there are related issues), others are introduced by the generation of the XSD from the model

Merged into Issue 15212

All services defined underneath “RmsServices” need to have their service operations detailed. They are in the PIM Model, the PSM WSDL model, and in the PSM WSDL artifacts, but not in the PIM service section of the textual specification

Operation definitions have been added

It should be clear that an attribute in the PSM represents a reference to another object and just which object-type it is. For example the attribute might be named a concatenation of the “rolename”, “_”, “ID” to indicate the referenced class via the role of the association. The type of the attribute would be IDREF

Resolved as part of Issue 15212. The convention eventually adopted involved W3C type NCName rather than IDREF.
Revised Text: None

Removing the ports will resolve problems of untyped ports in the XMI.

These two ports were remnants left from services that were eliminated and were therefore untyped. Furthermore, the remaining ports created tool exchange problems. The port diagram is only used for exposition to demonstrate the scope of service provision and consumption, and will be retained in the specification pictorially to provide overview, but they are removed from the XMI. The diagram is updated with one that correctly presents our current suite of services.
Additionally
1. There was a large quantity of Sparx Systems EA extensions in the XMI. These were removed.
2. The original XMI had PSM constructs included that resulted in portability problems. The XMI is now strictly the PIM and now loads cleanly into both MagicDraw and Sparx EA tools.

RecordCreator and RecordKeeper must have Agency, following the same rules/pattern as Provenance designation. The ultimate organization of the RecordCreator and RecordKeeper must be the same.

[JRMS Remaining Issue]

It is agree this Issue is incorrect. It is quite possible for the RecordKeeper to be from an entirely different agency than that indicated by the Provenance of the record (e.g., when records are seized by a law enforcement agency... they become the RecordKeeper for the period of time that they have custody of the record.)
Revised Text: None

On deletion of a record from the RM system what information is kept to track that it existed and has been destroyed and by what authority and perhaps accessioned by NARA or other Transfer.

[JRMS Remaining Issue]

None. However, there may be an additional issue to be raised as to what is deleted when the decision to delete a ManagedRecord is made. The model needs to be notated from ManagedRecord on out with aggregation or containment as to what is contained as part of the record and therefore what will be removed when the ManagedRecord instance is removed.
In terms of what information is retained (as addressed in this issue), that is dependent on the applicable business rules and regulations and therefore is out of scope of this specification.
Revised Text: None.

Package: UseCaseRealizations –Find Disposition Candidates Realization. This package was removed from the submission due to incompleteness & should be added back in and completed.

[JRMS Remaining Issue]

Out of Scope
Revised Text:
None

The specification should have a paragraph that lists all the RMS's mentioined in the UFAC spec, and DoD 5015.2. AS 4390 Australian Records Standard, and all other non-european standards, ahem, not mentioned in the spec.

A paragraph above them, simple and sweet that explains why the OMG RMS specification is compatible with them, listing all these current standards and specifications.

[JRMS Remaining Issue]

Out of Scope
Revised Text:
None

Package: RmsUseCases

[JRMS Remaining Issue]

Out of Scope
Revised Text:
None

According to the Spec, a managed record can have many categories and a category can have many disposition instructions. Therefore, a managed record assigned multiple categories can have multiple dispositions unless a specific user action is taken to assign one disposition plan for the record set that the managed record belongs to. Please explain?

The resolution to the multiple categories is in Section 8.2.4.3 "Category::RecordCategoryAssociation" where the association between ManagedRecord and RecordCategoryAssociation is documented (emphasis added). The multiple RecordCategoryAssociation's are necessary to keep the history via the previous/next association. There is an analogous description for a Category's multiple DispositionInstruction's (only one is current... the remainder are historical)
Revised Text: None

The XMI is not well formed & does not play well with other tools. This problem may correct itself since EA is participating in the XMI interchange working group. The original XMI was generated with V7.1. (MPG used a much later version). We will keep our tools up to date and stay in communication with Sparx to see if this issue can be resolved

For AttributeProfiles it is important to distinguish between user entered and system entered fields.

This is a concept that has been an important design criteria for RMS from the beginning.

We must assure that all fields and operations for accomplishing this are provided.

Many of our operations operate on one object at a time. For efficiency a single call to perform the same action, or to provide multiple instances of object values needs to be accommodated. (e.g., return sets of id's, set many attribute values on an attributable object at one time, set-aside many records at a time.)

For example “type” should be “attributableClassType”; all the “definition” (4 of them) should be “attributeDefinition

Using plain Xquery can introduce security issues when it comes to authentications and authorization. Has this approach been revised?

Our current service model often requires numerous operations to accomplish a task which must be completed or else the repository will be inconsistent. The task of setting aside a record, for example, is not a single operational call. Should connection be lost between the client server, or any operation fail, the repository can be left in an inconsistent state. This violates one of RMS original design principles

Query Service mentions that input parameter qualifies the requested elements and the return string contains the elements that match the request. The Spec does not define a schema for the query string.

The disposition plan is executed against a record set,
but the spec does not define how managed records are added to a record set. Is there some insight into the expectation?

To fit the broadest spectrum of business processes we need to assure that only the absolutely necessary attributes for records management are required.

Issue identified in the San Antonio TC RMS-FTF Meeting

The CaseFilePart description does not properly reference CaseFilePartDefinition concerning constraints on the CaseFilePart

Managed Record should just have an association to Party to show which party created the record.

This will determine the manner in which Sparx EA will traverse the UML graph to produce the XSD

The link is needed in order to require attribution based on DocumentType; which is the intent

We list a party ID as an attribute in the style of a foreign key in the PIM. Generally we explicitly model associations in the PIM. It would however remain a foreign key in the PSM XSD.

We presume global uniqueness in our ID scheme. Is that assured by the W3C type or do we require further clarification of intent of identifiers.

For example, an AttributeClassType can have many DataProfileAttrDefn’s (maxoccurs=”unbounded”, which is the default).

document.id type is “string”, it should be “ID”.

Description: This defines the number of times a certain case file part can occur in a context (the case file or a case file part – once hierarchy is supported). Example: an auto damage claim case, whereby four photo’s would be required, e.g. one from each side of the car. So, cardinality (multiplicity) of part “Car body photo” would be 4 (associated to the corresponding case file part definition).

Description: This allows referencing one case file part from another, potentially in another case file. This would allow referencing e.g. a customer from an order.

Package: RecordCreatorsService

[JRMS Remaining Issue]

A practical system requires an effective exception architecture.

Nominally 80% of the normal operation of a given system is in exception processing, not to be confused with related concepts of system errors or failure.

Need to devise an exception architecture suitable for informing user of errors and to allow developers to debug unexpected responses.

[JRMS Remaining Issue]

We must state that a hold on record attributes must disallow changes or updates of any sort, or enumerate the specific changes that are allowed, if any.

[JRMS Remaining Issue]

The current discussion of Record Schedule and its relationship to categories is inadequate.

[JRMS Remaining Issue]

Package: CategoriesService: Do we need Ids for CategorySchemas and RecordCategories? Is Name unique or the path down the form the top to the record category?

The presumed concept of RMS operation is that most communication will be done between the client & server via the ID's of the objects being referenced. We need to assure we have ID's on all entities that require it.

[JRMS Remaining Issue]

Description: This enables building a structure like an order with embedded lines. Example: Sales order for professional product, e.g. aircraft (order from e.g. Continental to Boeing). There’s a lot of “documents”, but there’s also a lot of “typical application data”, such as in CRM, ERP, etc. systems. It should be possible that case files refer to both in combination.

The FTF needs to use the appropriate namespace URLs for the XML schemas and WSDL documents associated with the RMS spec. The current draft of the SMSC format for these URIs is at: http://www.omg.org/members/cgi-bin/doc?smsc/09-06-03.pdf

How many In Force authentication methods can there be? Does there have to be at least one in force at all times?

[JRMS Remaining Issue]

Description: Case file parts are attributable. To express structured data, you not always need an associated document. It should be possible to explicitly mark a case file part as 'no document expected'.

As it is currently designed, the RecordPart requires the content to be stored as a hexBinary blob. This is very convenient to deal with non-XML data, but it's considerably limiting when dealing with XML data.

Considering the emergence of XML databases, if the content of the record part were stored as regular XML, I could perform a XQuery statement that could traverse both the RMS-related data as well as the record-specific data.

This is also aligned with the emergence of XMLSec (allowing digital signatures embedded in XML documents).

The simplest way to solve this issue (while providing backward compatibility) would be to provide a choice between "content" - which would still be the hexBinary blob - and a new "xmlContent" element - which would then be of the type "any" and then could store any arbitrary XML document.

There is probably a need to consider language talking about the inherent conflict arising from the proper use of terms in the OMG RMS Spec and the use those same words have in the DoD 5015.2 Standard. For example, the word "Agency" is specified in the OMG Spec but it is addressed differently in the 5015.02. The 5015.02 operates more from the individual/action officer level with regard to this and the OMG Spec applies itself to satisfying 44 U.S.C. and 36 C.F.R. (Daryll Prescott, 20091207)

Among other things, Annotations are used to tag ManagedRecord's with their security designations. There are situations in which only some RecordPart's are classified, therefore we need to be able to annotate RecordPart's separately. There was discussion of allowing Annotation's to Document's directly, but it was pointed out that there are situations in which a Document is not classified, but in the presence of another Document it is.

DoD 5015.02 does not specify field sizes. Nor does RMS in most cases. The description of a photograph might be adequately handled by 500 characters in one agency whereas another agency may require 50K characters. If the size is specified, what happens to inter-operability?

There needs to be a double delete authority, two seperate persons (personalities) in the system before something that is identified as a record can be dispositioned. (Daryll Prescott, 20091203)

Would a caseDefinition, as core object in case management extend caseFileDefinition, or associate with it? (Henk de Man, 20090225)

Name: Daniel Ruoso
Company: Prefeitura de Fortaleza
mailFrom: daniel@ruoso.com
Notification: Yes
Specification: RMS
For the fields effectiveStartDate and effectiveEndDate in Role, that information might not be known at the time the record is first captured. Are these fields required? If not, how do I work around that?

We need to require that it be done on capture. Should we require other events like retrieval? (This would get into the business process of the organization

The RM environment when an authorized user wants to "open" a record, a copy of the ManagedRecord must be provided to the user to open, or placed in a temp directory to open to preclude any possibility of any change

We need to provide the capability to include/exclude records that are on hold for Disposition Candidates identification. (Bill Manago, 20091204)

In the specification the package diagram does not include, for example, "Disposition". (Larry Johnson, 20091209

Clarify that we are using the steriotypes from SOAML

Title: RecordKeepers and Suspensions Services no longer exist but their ports are in the specification and the XMI file.

Descr: Removing the ports will resolve problems of untyped ports in the XMI.

RMS-FTF: We have the solution for the issue (discovered in Monday’s AB Meeting) and will apply it. A ballot will be issued as soon as we verify that we have clean XMI, and again, I will be asking for a quick turnaround. We need to have this fixed to go through the AB on Thursday

The XSD makes properties all lowercase as opposed to the camelcase from the PIM, which makes them a lot less legible and inconsistent with normal XML Schema practice and inconsistent with the names generated for ids. For example effectiveStartDate becomes effectivestartdate. But we also have hasRoleID which is camelcase.

In the XSD, references to other elements have been incorrectly typed as xs:NCName rather than xs:IDREF

Many elements are in the XMI as primitive types which are not at all primitive, for example RecordKeeper, ManagedRecord, PartyRoles[], Party[], SuspensionRevocationMethod, AuthenticationMethod, ActionEventSpecification, DispositionSuspend, Revocation. Some of these (e.g. DispositionSuspend) have both a class and a primitive type. I suspect this is because they have been used in operation parameters.

The XMI references several Sparx-specific PrimitiveTypes in packages EA_none_Types_Package and EA_PrimitiveTypes_Package

Typo “subordintate” in 8.4.4. and Figure 8.5 and the XMI and XSD