Friday, April 30, 2010

The software industry is broken

Too many cooks and not enough broth. Doing something simple involves days and weeks of discovering the myriad of ways that developers define simple.

Each engineer wants to impose some new intuitions when the previous intuitions are just fine. Openoffice is plagued with the problem, as was SUN; and they have little market share. Microsoft gets it, eventually, and controls proliferation. But they charge a large price for keeping their engineers under control.

Engineers, especially the ones who consider themselves very clever, will eventually lose their jobs as software learns how to avoid their cleverness. Trying to do one simple thing becomes a task of avoiding the extra "mal-investment" created in the slightly more complex.

The business has becomes a business of sorting out the "too clever" from "the just about right"


So then one tries to register for one of the help sites, but the login is too complex and there are too many competing software help sites. The problem seems to compound exponentially. So we end up with one user, twelve software packages and 12 help web sites, leading to 144 choices for the single user. The result is that it is often easier to write qa new application from components then sort through the myriad of software applications and support on the web.

All the executives in the software business know why Microsoft won this game; two or three days of fiddling with all the developers and their variety yield $600 in costs, enough to buy the Microsoft product.

This all started by a simple effort manipulate simple data with SQL. That led to either working in dos with SQLite, or getting a small GIU for SQLite and navigating the myriad of web help trheads. After a day of navigating the web, I selected DOS. Now I do not remember any of the DOS shell commands, nor do I trust Microsoft to have presered the DOS environment!

It is like being forced back 30 years and finding the stuff that used to work no longer does, and the current stuff is written by developers who design to confuse.

But in the end.

I got it all working with SQLite, the little engine that reads big BLS prefix code files. Then SQLiteStudio does the trick. I simply need to sort through codes and get time series from BLS. Then they go into dataframe format to R, and bingo, can tell you who did what when.

No comments: