Analyze file type rather than guess from filename

Would it possible to have a better gauge on the correct mime-codec for MediaSource APIs if you analysed the file? There are tools are tools available such as mp4box.js, and we _should_ only need the first few bytes to figure this out. Could we read enough data from the file's readstream and use that to get the full codec data before piping that and all the rest of the data to the writeStream for MediaSource?