Message-oriented applications involve an interface defined around messages (or documents). I've been working on applications that involve a fair amount of manipulation of XML and X12. Recently I started yet another, and started to recognize the essential design patterns.

There are four use cases. The first two use cases define …

Some recent thoughts on using Python, directly, as configuration file syntax. In a way, a configuration is a highly-specialized Domain Specific Language focused on the application's problem domain. The point is to leverage syntax we already have to create meaningful configuration files.

"Python is sometimes a good use for a …

Here are a few more Passive-Aggressive argument techniques that can derail a project. Rather than provide an alternative or improvement, this passive kind of aggression raises things which are "potential" problems or issues. Things which serve only to raise doubts.

This approach plays into the hands of risk-averse management. It …

Here's one recent example of a significant bug.

The code said A is not NULL OrElse A != " ", which doesn't do anything useful. If A has non-NULL data, this is true. The second part of the clause (A != " ") is only evaluated if A is NULL. If A is NULL, no other …

I have to say this first -- just because ERwin calls it "physical" that doesn't mean anything. ERwin uses "physical" model to mean "product-specific logical" model. They use "logical" to mean "product-independent logical" model.

ERwin doesn't do physical modeling. Not even a little bit. The physical layer of a relational database …

TC's full quote was "We consume a significant portion of the overall contract generating a functional specification whose sole use is to serve as the outline for the acceptance test. Those who choose to argue this point might like to explain the recurring customer comment 'You did an excellent job …

TC Writes "With 30 years in the biz, I’ve experienced all the pains associated with a rigid linear approach (functional spec -> design doc -> code -> integrate -> acceptance test, always being sure to maintain a current version of the traceability matrix). An iterative approach is inherently more appealing, but I’m …

Time is simple. You can make it complicated with any one of a large number of dumb-as-dirt decisions.

In the Real World™, time may be a quantum field (read stuff by 't Hooft ). For that matter, space itself may have some quantum granularity. Or, this could be a handy hack …

The money quote was this "It’s not a big leap then for me to consolidate all three of those into one uber relationship table...".

The question was a good one because it involved the classic higher-order database relationship.

The canonical SQL relationships are binary equi-join relationships that can be …

There aren't too many variations on the open-ended date range problem. We have a table of events or rules or policies that have start dates. Some have end dates, because they've been replaced or superseded. Others don't have end-dates because they apply for the foreseeable future.

There are two kinds …