• Web version aimed at neutralising the threat from Google spreadsheets.
• More integration with Sharepoint (whatever that means).
• Better data manipulation facilities.
• Sparkline graphs.

The things that worry me about Excel won't get fixed for years I am sure: proper 3-dimensional formulae; sensible evolution of its function library for business; the future of VBA and whatever might succeed it; making it easier to build addins in compiled languages. Oh well.

One of the main questions is whether to go for a cardfile type simple list or fully normalize.

On the face of it, it is a technical no-brainer. But if you cogitate on this long enough, it becomes quite a teaser. When, for example, does a list of widgets actually become a list of components of type widget or a list of things of type component of type widget. Before long, you can find yourself writing the database of everything. And it takes longer than the beautifully simple list.

The pros of normalisation are in every textbook. The cons of normalisation are often not. Here are what I think are the cons:

• Your input forms become harder to write. Adding an attachment to widget involves adding an item to attachments and linking it, instead of simple filling in the items 'attachment1' field.
• Retrieval will involve one or more complex queries with joins, instead of just sucking in the whole table.
• Users will find it harder to understand the internal structure of the database - they may sometimes get confused.
• Quite often, a denormalised table is what is required as an output, because we like to look at long lists with all the fields showing. I found that if you import to a spreadsheet Excel you nearly always end up denormalising because you want a single table sucked in and Excel is not relational like a database is.

JD's normalisation criteria:

1. If in doubt normalise because its the RIGHT thing to do.
2. If all the outputs of this database require a denormalised output and/or your data inputting users would do their job much quicker with denormalised, then go for the simplest flat list you can.

I've just completed two Brainbench certification tests on Microsoft Excel 2003. I passed both, one at a Masters level, beating in one case 97% of other test-takets and in in the other 99%. These are by far the best scores I have got on any Brainbench tests, which I have a fair degree of faith in.

But here's the thing. On the second test, I was guessing more than half the time. I mean, I really didn't have a clue - I have never used Pivot Tables, OLAPs or XML - I just don't believe in them.

So if I don't know what I'm doing and I'm better than nearly everyone, most people are totally lost!