I have been shooting since February 2004 (writing this in August), and not really a lot. At the moment my photo library is around 2Gig, so it is crucial that there is some sort of system to the madness. Here’s what I have done:

• My Canon G5 is set to "no counter reset". This will mean that every image automagically gets a unique ID.
• When I download images, I put them in a folder structure "year"/"month"/"day"-"topic" (think backslashes if you're on Windows). So the directory 2004/08/03-gerbils (or 2004\08\03-gerbils for Windows) will contain pictures I shot of my Gerbils shot August 3, 2004.
• When I start editing, I put the work-in-progress in a subfolder which is called 'work' (2004/08/03-gerbils/work). This keeps the originals ('negatives', if you like) from the modifications ('prints') and allows me to selectively backup/ignore either kind of image.
• When I have something finished, I move the final image file to a 'final' subdirectory (2004/08/03-gerbils/final) and create a high-quality jpeg export for web publishing.
• Then I publish worthwhile stuff using Gallery Remote to my Gallery site
</UL> Snapshots or picture series about parties, etcetera, are captured in JPEG on the camera, downloaded, and uploaded to Gallery in batch; usually I mail the party members a password that will allow them to access the full-sized images so they don't bother me for reprints.

If you want to find images again, it is nice to have some sort of image cataloguing program. It seems that IMatch gives you a lot of bang for the buck here - with IMatch, you can attach structured data to photos (what was photographed, and were, and whether the picture is any good, etcetera) and subsequently find stuff through that data or even through the actual picture contents (it has a 'find similar pictures'). As it keeps stuf in a separate database, you can move the actual images off-line (to CD or DVD backup) and IMatch will still be able to locate them.