This discussion has also been overly focused on what a data dictionary, vocabulary, ontology, etc. is, and hasn’t paid any attention to the issue of how one knows or specifies what such a construct covers – and doesn’t cover. That goes to the heart of your point about interoperability across silos. Removing “systems” from the equation, or talking about “de-silofication” doesn’t change the fact that their data has scope similar to, if not identical to, that of the scope of the system, silo, domain, or enterprise for/by which the data was created. So one still needs to know what a given data dictionary, vocabulary, or ontology covers, and doesn’t – in addition to the issue of what types of information such constructs might contain about said coverage scope. Then we can start talking about which particular systems/silos/domains/enterprises need to interoperate in order to achieve some desired objective(s), and what the ontological “impedance mismatches” are that might need to be overcome between particular data sources (over and above any system-specific implementation mismatches and “information hiding” features of specific systems that Kingsley keeps complaining about). Note that this is also the main issue with ontology reuse, albeit in a more narrowly constrained sense. i.e., are the impedance mismatches small/unimportant enough to ignore (or work around) for the reuse purposes at hand.
The reason there are few “distributed data dictionaries” is because that would require some enterprise/business/governance model that spans the scope of multiple such silos/enterprises/domains. Typically, that has been the province of industry domain consortia, like the Open Group effort to develop “business objects”, and third-party information brokers, such as travel web sites, who have to deal with a multiplicity of domains and multiple enterprises within each of those domains (hotels, rental cars, airlines, tour groups, buses, ships of various types, customs rules, etc.).
From: ontolog-forum-bounces@xxxxxxxxxxxxxxxx [mailto:ontolog-forum-bounces@xxxxxxxxxxxxxxxx] On Behalf Of David Eddy
Sent: Friday, February 14, 2014 8:20 PM
Subject: Re: [ontolog-forum] What the difference re., Data Dictionary, Ontology, and Vocabulary?
On Feb 14, 2014, at 7:47 PM, Kingsley Idehen wrote:
I have spent more than 20+ years of total dedication to making new and emerging technologies work with existing (so called legacy) systems. I founded OpenLink Software to enable integration of data across artificial data silos, created by *myopic* applications.
But that's the DATA the systems produce.
I trying to talk about the SYSTEMS, that produce the data.
The SYSTEMS & the DATA are not the same thing.
The SYSTEMS are the machine tools that produce the end product, DATA.
I am reminded of wisdom from the 1840s when industrial America was learning how to make things. It was noticed that building quality into the manufacturing process is far more efficient than inspecting defects out.
Just handling the data—the manufactured product—is an exercise in futility until a firm understanding of the upstream manufacturing process is fully understood.
If one doesn't know which systems, programs, logic, data structures & rules are producing the data, how will one know when the data suddenly changes?
The sort of data dictionary product I'm talking of—decidedly not a list of data elements—enables organizations to do impact analysis—what's connected to what—so that we're not constantly repairing down-stream errors.
integration of data across artificial data silos,
Integration is not possible with silos. You mean interoperable. There's a huge difference.
Integration is possible when one is in command, owns the budget & has a blank sheet of paper, otherwise the only option is interoperability.
The silos are not necessarily artificial. They were built that way for valid reasons... available skills, limits of technology, budget, vision, deadlines, organizational boundaries, etc.
Silos are a complex reality we're going to have to learn to deal with... as John Sowa points out, these systems are going to be with us for decades.
Understanding language & meaning across Silos would be an extremely useful application for Ontologies if they can be commercialized.