Although this is rather an old thread, which would make me more of a Thread Necromancer, I just spent 10 minutes converting the 17 BMPs from one of my forts which used a graphics set and a fancy graphical font (16x16) into PNGs using Paint.NET, and it compressed them from 1 GB to 9.78 MB, and it took me 10 minutes to do it, which took much longer than using the (er, my) Map Compressor. Now if I could tell paint.net to batch convert all the files that would be easier, but it still would have taken at least 5 minutes, I think. Anyway, there are programs that WOULD batch convert them, but then you have the hassle of getting them, etc etc. Onward:
PNGs: 9.78 MB.
Now, X said that an order of magnitude improvement in efficiency wasn't good enough.
So, I compared it to my fdf-map from the same images: 145 KB. Every time you have an image of a plump helmet, in every map, that image will still only be stored in the fdf-map file *once*. Especially with the multiple maps, there's just no way for conventional image compression to take advantage of the images being in multiple files because there's no way for them to analyze multiple files and identify patterns common to all of them (especially without running out of memory).
fdf-map: 145 KB.
For those, the fdf-map file is 1.4434% of the size of the PNG files, which is just about two orders of magnitude now, took less time to make, and was easier to make. The only problem would be that you need .net 2.0 or mono, and mono doesn't seem to work everywhere, and although vista supposedly comes with .net 2.0, apparently a few people have had trouble with that (I saw some forum posts by this one guy)? I've got links right to where to get it in the page(s) for downloading the compressor though.
Someday I'd like to rewrite/port the map compressor in another language which is portable and doesn't require additional stuff to be installed by the user, rewriting it as a portable library which DF can call functions in to directly write out an fdf-map file, rather than writing out BMPs. Then it could be included with DF, and users wouldn't have to do anything except hit export from inside DF, and upload their fdf-map file to the map archive. But currently that's not a priority. (I do know several other languages, including c++, but would probably end up attempting it in haskell instead)
I'd like to be doing something similar to what Toady's doing, making a good detailed free game, with modular text/tile-based graphics, and actually earning a living from the donations (eventually), rather than not making any money and not working at anything I like, and that (albeit distant) goal is the main reason the map compressor library rewrite idea is basically just an idea and nothing more at this point. (It's also the same reason I suspended pretty much all of my other projects)
If at some point in the future I've got time to devote to side projects and decent income, I'll probably revive the idea.