Kyle Banerjee wrote: > Starting with data modeling is like trying to learn a new spoken language > by focusing on grammar [...] Hmm. It seems that a lot of people are, shall we say, somewhat misguided to what data modelling is, even mighty WikiPedia who makes it into a formal process of sorts, and I can see it repeated ad nauseum wherever you go, giving us the idea that it is all about the schema of columns and the nuts and bolts of tables and relations in a RDBMS. That's confusing data modelling tools or processes with the generic open-ended category of data modelling. Data modelling is simply the act of exploring data-oriented structures. Over time I've learned that everything we do, every little problem you battle with in your every daily life, revolves around some data structure, the names of such, and their internal and external relationships. The simplest web form has a model, simple and complex applications do as well, enterprise systems, library systems, formats, databases, documents, spreadsheets, this conversation, your bicycle, your morning routine, *everything*. There is, in my strong opinion, a horrible conflation of the concept of modelling data and implementations pinning down data types; it's an evil so strong it blinds us, cripple us, and I feel like screaming out in terrifying agony the horrors within! The wrongly applied indeces! The labels on columns! The semantic binding of one sub-structure to another! The optimising tricks used! The stored procedures! The conceptual semantics of labels in n-ary graphs!! *aaaaaaaaaaaaaaaaaaaaaaarghhh!!!!!!* The wretched *name* of a single field and how it quietly eats up any disambiguous notion we put in place, through the many well-meaning but afflicting layers of abstraction and implementation, it drives me insane! "Name"!? What does that mean in the context of an email address? What does "comment" mean when it reaches my ORB? What were they thinking when the model designed resulted in SQL statements 1K long? There's so much information written of the topic of data modelling, and most of it ignore that very thing that it should embrace and focus heavily on; good semantic design. (Granted, it has become far more focused on in the last 10 years, and I'm extremely happy for that) Put some heavy thought into your tables, because what you perceive as a simple table of users becomes an overwhelming problem when you add special users to the system. Have any of you ever created an ILS with a table "book" in it? (C'mon, raise your hand, I know you have!) Yeah, that's the sort of evil I'm talking about! Libraries don't deal with "books", they deal with bibliographic meta data of objects, and sometimes those objects are called a "book" which has certain constraints and properties that link to special meta data that isn't static. Version 1.0 of any system if famously rubbish because of the learning process of getting all this stuff wrong. Version 2.0 is famous for being overly abstracted and incomprehensible. Version 3.0 is getting there, but you're bogged down in the middleware, translating between good but incompatible models. By the time you get to version 4.0 you realize that the underlying concepts which drove versions 1 through 3 are flawed, and you need to work in terms of FRBR sub-graphs instead of MARC records. Version 5.0 is so re-written and re-conceptualized, you decide to call it something else, version 1.0 And we repeat the cycle. If your software isn't like this, consider yourself lucky (or at worst, self-deluded :). > Data modeling is extremely useful, but > mistaking drips and drabs of it early on for reality can poison your > thinking. Sorry, you got that back to front. We all agree that understanding what user want and / or need is King, but unless you've got that understanding of not only what the users want but how systems can deliver this without creating constraints that will screw things up when you extend that original delivery idea, you're going to suffer. Badly. It's easy; take great care to what you call things in your system (no matter whether it's in the database, your objects / classes / instances / interfaces, user interface, buttons, messages, windows, data types, loops ... they're all data models that need to be as cooperative as possible, speaking the *same language*, to be compatible in the meaning they give the concepts used. If your Wheels API has different semantics from your Steering API, making that car is going to be a really crappy experience, for you as a developer, for testers, for maintenance guys, for service people, and most of all don't think for a second that the driver won't notice. These semantics are far more important than what our industry traditionally have given them, and in my opinion it is our biggest flaw. Trust me, I've stared at data models up and down so many systems over the years (10 of them in a high-flying big consultant agency where we came in when projects otherwise failed) it's amazing I'm still sane. But just like in the Cthulhu myth, little by little I feel my sanity waning from staring at all those hundreds of assumed harmless semantic differences and oddities in them, the little things that at first looks like it possibly wouldn't cause any kerfuffle. Those are always exactly the little gremlins that cause big hickups further down the line. None of these little engulfing little smiles of hell should be kept alive. Just like two little cute rabbits in a meadow is adorable at first, you curse yourself for bringing in the rabbits to this land where they have no natural predators, and they cause an expensive and horrible mess that sometimes are simply impossible to clean up. All this from just two fluffy rabbits. Well, I say; don't trust the rabbits! They're all evil rabbits of Caerbannog! Kill them, kill them now, with grenades if needs be ... *ahem* Where was I? Ah yes, evil little gremlins. You think them innocent at first, but they will destroy your otherwise carefully crafted systems, make you cry and question your worth. Don't let this happen to you! Don't think that pure programming can even fix this problem, because it's a very human problem. It's human abstract thinking that fails, and pollutes digital logical systems because something wasn't thought out well. So, young Padawan, may I suggest you understand this before you leap forth in haste, armed with information and programming rather than knowledge and understanding of semantics? > The most important question to understand is why because it drives > everything else. No argument there. For example, why are we having this conversation? ;) Regards, Alex -- Project Wrangler, SOA, Information Alchemist, UX, RESTafarian, Topic Maps --- http://shelter.nu/blog/ ---------------------------------------------- ------------------ http://www.google.com/profiles/alexander.johannesen ---