OK, my last post said that my latest "pie in the sky" idea that I probably won't implement, is pie in the sky. To clear things up, that's not because of problems with the CRM Discriminator. It's not even that big an issue to get a space to hold all that information (but paying for the space can be an issue). It's just that being able to search through information quickly is very difficult.

We're used to Google giving search results almost instantly, and Amazon giving us datamining results before we even ask for them, so many people don't realize this. Google and Amazon were some of the first companies to need databases that do more than what database companies design them for. Your normal database is organized so that you can pull up, say, an order history based on a customer's name, or a particular order based on the order number or date. But if you say "I'd like a list of all the customers who ordered product X," the database will have to actually look at each order one-at-a-time to give you the answer.

There's a lot of interesting research going into this, but so far you can't buy the right database for this kind of work. You've got to roll your own.

UPDATE OK, it's not pie in the sky. It's Findory, and after a little looking around, they don't have the kind of readership I'd want for good recommendations.

So, back to my "what can I do with CRM" puzzle.

UPDATE 2 A better description of the database problem.

If you haven't already watched Dr. Strangelove or: How I Learned to Stop Worrying and Love the Bomb, it's a good movie. It's also the source for the name of a tool I'm in love with, but can't find a use for, the CRM 114 Discriminator.

OK, I know it's a great tool for spam filtering. I may use it for classifying email at some time. However, I know there's a lot of power in there, and I want a use for it. Something like a cross between Slashdot and Amazon's recommendations system. That's what I thought Reddit was, but Reddit just goes with popularity, not "popularity among people who vote like you," which is what I'm thinking about.

But for that to work, I'd need a lot of readers. Not just a lot of readers, but a lot of diverse groups of readers. That is, enough readers that people who like Hemmingway don't get lumped into a general "readers" category with people who like Tad Williams (and Tad's prolific, so his group should probably be broken down into "Tailchaser's fans," "Memory Sorrow and Thorn fans," etc.). And "programmers" should be broken up in "Ruby fans," "Perl fans," "Java fans," "C++ fans," "CRM 114 fans," etc.

Oh, and I'd need a place to put all that information. And, while I'm at it, I'd like a pony.

Unless somebody else can think of a use for the CRM Discriminator.