Windows Journalist Takes On Tiger
I guess from my point of view as a coder, determining the visible text strings of an entire PDF vectored image might be more complicated that looking up the meta-data for an image, MP3, or whatever, which presumably is nicely organized in either a static header or as a series of tag/data elements.
I'm a coder too, although I don't do any OS X development to speak of (although I am an OS X user). I do know, though, that the version of Preview in Panther already does an excellent job of text recognition, and so I assumed that Spotlight would use the same library to do so.
Grabbing EXIF metadata from image files is easy... what's hard is having more meaningful metadata than what camera took the picture and at what time. Searching for images having to do with "Niagra Falls Vacation" is a much harder problem than searching for images "Taken with Canon EOS 20D between April 1 and April 15, 2005". For that matter it's also harder than searching PDFs and Word documents for "Niagra Falls".
The real question I have about Spotlight is how meaningful metadata is set, not how it's searched. How is the "Niagra Falls Vacation" keyphrase getting set and saved in the image?