A metadata rant

Repurposed from a dm-discuss post.

Let me be a little contrarian and argue first (as head of the Metadata Management Office at a Fortune 100 company, and the manager of a modern MOF-based metadata repository) that

the terms "metadata" and "repository" are quite useless and need to be left behind.

Really, what we have is the overall functional area of IT or IS (take your pick). This general organization capability, just like its counterparts in Finance or HR or Manufacturing, has processes and data elements it needs to manage. It manages the definition of data, process and function models, the physical data and software artifacts implementing them, hardware computing platforms supporting those artifacts, and a host of process-oriented abstractions: change and incident tickets, work orders, services and systems as cooperatively defined with the client, and much more. It also manages the human and financial resources necessary to support the IT capability.

If this seems misguided, consider the following questions:

The definition of an "entity" would generally be accepted as "metadata." Is the name of the analyst who defined that entity metadata? Many tools can and do store it. What about the project in which the model was created? What about the financials underlying that project? The software quality practices? Were inspections carried out on the data model?

A logical evolutionary step for metadata repositories was to extend their data dictionaries to include the programs that accessed the various data elements. But are the change tickets that put those programs into production metadata? Are incident tickets related to those programs metadata? The headcount and budget required to maintain the system in production ongoing?

If you say no to my questions, then I have to ask: where is the ROI on your product? I can tell you what my clients are asking for: an integrated picture of the IT environment, with complete transparency and traceability across those questions I'm asking.

You can see the problems quite clearly in the CWM, which has "Email" and "Telephone" as metamodel entities, which has been controversial on the dm-discuss list. I support the pragmatic argument that IT assets are created by people we might want to reach, and therefore their contact information belongs with the asset metadata -- but the results do seem curious, as contact information becomes both metadata and data -- it "crosses levels." Hence my argument that it's all really just "ERP for IT."

Regards,

Charlie

Comments

[David Eddy]

Charlie -

Good rant...

>
A logical evolutionary step for metadata repositories was to extend their data dictionaries to include the programs that accessed the various data elements.
>

Ummm... errrr... ahhhh...????

"A logical evolutionary step....was to extend their data dictionaries."

To include the programs & contained there-in data elements (actually system contains program contains copybook contains data element) is the STARTING point of any sort of tool that could reasonably claim to be a data dictionary.

I am explicitly excluding from my definition of data dictionary a simple listing of data elements & their meanings.

Me thinks that the basic understanding of what "data dictionary" actually means (as used in practice) to any particular organization.

...but such is this cloudy discussion... the words can mean just about anything you (or particularly the salesman) want, without actually challenging someone for their definition.

- David

[Jessy Jame]

Can’t you just say what you mean? I don’t really understand all this. But I like the way you do it. I can surely learn a lot on your sites.

[R. Todd Stephens]

Perhaps another question to ask is... Why do other technologies, frameworks, and methodologies evolve over time and metadata does not? Who is responsible for holding metadata and repository definitions back in the 70's? Why is the term registry better than repository? Hell, let's stop calling these things personal computers, that was the commodor 64, Texas Instruments TRS-80, and the IBM Dual 5 1/4 drive PC... Evolution occurs, Metadata has moved on... It's time we all do the same. :-}

[Shawny Diane]

A friend told me of your site. That´s definitely what i was looking for. I will surely recommend you.

[DaveHay]

Charlie, I appreciate your rant. I had come to roughly the same conclusion when I realized that, while I can go into a company in pretty much any industry and, thanks to data modeling, can very quickly understand the nature of its business, nobody had really done the same thing for the IT industry.

So, when I set out to model a metadata "repository", I took it as an assignment to model the IT business. My book describing the results, Data Model Patterns: A Metadata Map, should be out within a couple of months.

I used my version of the Zachman Framework to organize it, so you will see indeed that metadata is data about not only data, but also processes, people and organizations, locations, timing, and motivation.

I observed what you did, that when talking about people and organizations, for example, the model of PERSON and ORGANIZATION is exactly the same as in the business model. What is different is that here PARTIES are responsible not for manufacturing or accounting, but for defining entities, managing programs, and enforcing business rules (among others).

I didn't address the financial side of the industry, but then I often don't model that in business, either, since accounting is itself a model of the business.

Thanks for your views.

Dave Hay