I prefer using
SubRip to OCR the subtitles. Contrary to Bonzi's experience, I find the process goes fairly quickly and is quite reliable. After the SubRip process, you only have a 50-100k textfile to deal with. There are some other advantages to using text-based subtitles, besides the slightly smaller filesize. For one, you can resize the text or change the font as you please, which you can't do without OCR-ing the subtitles. For another, you can render the subtitles in a high-resolution, anti-aliased font, which looks much better than the bitmapped font that a non-OCR subtitle ripper gives you. Finally, if you're so inclined, you can make your own additions or changes to the subtitles.
Another reason why I use SubRip is because I always seem to have trouble remembering how to work VobSub.