RE: [REP] Data versioning

> The final trust example will also be a Repository use case, and will
include the requirements for data authenticity/validity/traceability.

In this particular requirement, the group has consistently recognised
that contributors of information may not be able to offer any guarantees
with respect to the validity of the data. Community contributions are a
particular case in point. In such a situation, we require that the API
will still support access to such data, but we must be able to discover
that the data has not been validated. I presume this is also part of the
"traceability" to which you refer.

Interesting point about the use case involving a property name change.
If the two names were in fact referring to the same thing, we could
consider this as an alias situation, and the solution would be trivial.
Though I suspect that people will live with the naming error. (Just
consider the case of the "referer" [sic] header in HTTP.

---Rotan.

-----Original Message-----
From: public-ddwg-request@w3.org [mailto:public-ddwg-request@w3.org] On
Behalf Of Smith, Kevin, VF-Group
Sent: 04 April 2007 11:32
To: public-ddwg@w3.org
Subject: RE: [REP] Data versioning


Thanks Rotan. 

Consistency is a use case which I can add to the Repository document.

Rollback should be a Repository requirement, and that implicitly
involves versioning.

Versioning applies to values, but maybe also property names? I can only
think that would be because of an error in the original name. As you
say, a user may decide to stick with the old name for backward
compatibility. If that happens then another complication is added:
should a query to the old property name return the value last assigned
to that property name, or the latest value assigned to the new property
name?

Versioning of the vocabulary would fall under the scope of that document
(although the Repository document may make reference to it).

The final trust example will also be a Repository use case, and will
include the requirements for data authenticity/validity/traceability.

Cheers
Kevin


-----Original Message-----
From: public-ddwg-request@w3.org [mailto:public-ddwg-request@w3.org] On
Behalf Of Rotan Hanrahan
Sent: 04 April 2007 11:09
To: public-ddwg@w3.org
Subject: RE: [REP] Data versioning


First thoughts....

Versioning is for consistency (I may decide to stick with a certain
version because I have evidence that it works for my business use cases
and I fear anything that might try to "correct" this).

Versioning is for rollback (if my fears are realized).

Versioning is for efficient replication/caching (so that I don't keep
making copies of what I already have).

Versioning applies to values (as the most likely changes in the bulk of
the data will occur because of corrections of past errors).

Versioning could apply to the vocabulary or ontology (because these will
evolve, not just grow, because we are not gods who can foresee all
possible future issues, no matter how hard we try).

The last one is something we should think about carefully. This could be
the most difficult issue regarding versioning.

Finally, the data in the DDR could come from multiple sources, even for
the same properties. For example, my company does extensive tests on
dozens of new devices every month, and so do my competitors. Naturally,
we expect our analysis to produce the same results, but this is the real
world and we know that any of us can make an error. We may make (a
subset of) our collection available (to the public) via a compliant DDR
API. So may our colleagues (a.k.a. competitors who also believe that
doing so will benefit everyone and is the right thing to do). This
produces an overlap in sources of information, but individual consumers
of the data can decide (based on trust, availability, efficiency, cost,
national allegiance, politics, favouritism etc.) to chose one source
over another. In a fair market, this is to be expected. The DDR should
permit this mode of operation.

---Rotan.


-----Original Message-----
From: public-ddwg-request@w3.org [mailto:public-ddwg-request@w3.org] On
Behalf Of Smith, Kevin, VF-Group
Sent: 04 April 2007 10:48
To: public-ddwg@w3.org
Subject: [REP] Data versioning


Hi all,

I have the following requirement in the Repository Requirements
document:

==== [DDR-DATA-VERSIONING] ====
(original text): The DDR MUST provide the capability to fulfil data
versioning, where data includes but is not limited to device
descriptions and their device properties.

(New text): The DDR MUST support versioning, either inherently or via
another system, of device descriptions. 

My questions are: How will this work in practice? Are there use cases
for having two or more valid/correct device profiles in existence for
the same device? Versioning implies a new value for a given property,
not a different set of properties. Or is this simply to provide a
rollback mechanism to a previous version?

Cheers
Kevin

Received on Wednesday, 4 April 2007 10:39:32 UTC