Changes in direction that the IT industry should have never made

From where the industry was when I started Dulcian back in 1995 to where it is today, I have seen massive changes in direction. Many of these changes make me miss the “good old days”.  I would like to point out some of the big shifts that were over-hyped and have not paid off in any meaningful way:

1. Reduced reliance on the database: I recently wrote an entire blog post on this.  Databases are useful, and to move away from them has killed more projects than any other factor.
2. COTS are better than custom builds: If your company is willing to reengineer itself to match what the COTS does and you keep customization to a minimum, then you have a chance of success. Twenty years ago, I heard my first COTS anecdote.  A insurance company paid $600,000 to buy a payroll package and then spent over $6M customizing it and getting it working. I have heard that same story retold over and over again. COTS are just as expensive, and  just as risky as any custom system design project.  The big difference is that the COTS company gets millions of dollars up front so once the projects get started, they are very hard to kill. My all-time favorite COTS story is from a developing nation that was talked into a $30M COTS financial project.  It is currently at $300M (and counting) with little success.  A senior government official told me (off the record) “I feel like I have invited the devil into my home and now I can’t get him to leave.” Publically, the country says only wonderful things about their very successful project.
3. Open source is best: We imagine that no can build a software product that has any value. If it isn’t a free, community supported product, it isn’t any good.  If this were such a great idea, we would have evolved into some kind of Darwinian consensus.  Instead, the industry remains more fragmented than it was ten years ago.  Look at any ten JavaEE projects and you will see ten different architectures.
4. 3NF DB design is not cool: When we moved away from COBOL in the 1980’s, we figured out over the next 10 years or so that normalized database designs are a pretty good idea. Non-normalized designs meant unmaintainable spaghetti code.  Now all the rage seems to be to ignore good design and build persistent copies of our Java classes.  And we wonder why these projects prototype so quickly but never deploy or scale?
5. Repositories are only for old people: If you store it in XML, you are allowed to keep metadata around, but putting it in a database is no longer cool.  At Dulcian, we put almost everything into some repository or another and then either use an existing code generator or write one for a specific project.  The more we rely on repositories, the faster and more reliably we build.
6. Business rules are now a fringe technology: We hit on the idea of using business rules thinking to drive all of our architecture. We separate the articulation of the rules from the implementation of those rules, providing a layer of abstraction that protects us from technology changes.  Much of the industry seems to think that business rules are only about process and therefore rules tools serve only to provide yet another place to hide your business rules.
7. The more tools you have in your architecture, the cooler it is: Fifteen years ago, a single developer could use a handful of tools and technologies to build an entire system.  Now, “it takes a village” and there are more moving parts than an Italian sports car.