I am from time to time reminded of the wonderful power of the spreadsheet: a big, blank magic piece of grid paper. Now Ingo says "Now the only part missing is some great new application, easy as Excel, running on the server...". Let me run with this a bit.

Spreadsheet cells are typeless: you can stick pretty much anything in a cell (in Excel, you can even put an OLE object into a cell). You can cast a cell's contents and reinterpret them. You can create cells whose value depends on other cells (i.e., their value is a function of some other cell/s). When you update a cell, the value of the function updates as well. That most programming languages don't work that way is probably the reason Ingo's step 7 is a meaningful part of the process, and in fact, is the point where the process fails.

I've been working with web services more as a client lately, and I'm really feeling the need for a way to handle this XML and SOAP stuff in some lightweight, extensible manner. XMLSpy and a few other editors give you a way to do this, but it's pretty brain-dead: write an envelope, send the request, munge through the response, extracting the data you need to produce the next request. And keep the result window open, because you might want to send several requests based on that response. That's better than the edit-compile-swear-debug cycle. Here's what I think a spread sheet for a web service would look like: I update the value of the "departure date" field, and I see the value of the "total air fare" field update after a few seconds. If InfoPath is in fact the enabling substrate for something like this, and if it's not tied to some other product that I don't necessarily want, then it's not just another software product, it's genuinely something to be excited about.

After reading Tim Bray's recent entry; on spam filters. Tim says "It won't be long before I start neglecting my occasional surveys of the "Junk" folder." and that rings absolutely true. I use Oddpost and Mozilla mail at home, and use the Python spambayes filter both at work and at home and I believe that I've started to take for granted how well it works. My work email has started to show some interesting effects: for example, the daily corporate newsletter suddenly got classified as spam, and now goes to the possibly spam folder. While this is pretty funny, it makes me wonder how that happened. I had thought that it's because I've become more agressive in classifying things as spam: I tend to get a lot of mail from signing up for software evals, and while that's not spam per se, it's generally stuff I don't care to read. In fact, the list of words that incriminate the newsletter could almost compose a spam by themselves: 'financial', 'grand', 'recovery', 'prices', 'stunning', 'sold', 'market', 'million', 'rights', 'stock', 'years', 'size', 'investment', 'buy', 'money'. Certainly legitimate in the context of telling employees about the company's performance, but otherwise damning. One thing that I'd like to be able to do with spambayes is to edit individual word probabilities. For instance, "travel", "tourism", and names of competitors should all be indicators that this is company business and not a financial scam. The filter should learn this eventually, but I'd like to be able to play with the parameters myself.