> Of course, you can also just reduce the Unicode pattern to bytes Ah, right. I was under the impression this was unsafe, since you could end up with spurious

> Of course, you can also just reduce the Unicode pattern to bytes Ah, right. I was under the impression this was unsafe, since you could end up with spurious byte matches that are not on character boundaries. But it seems the keyword is "self synchronizing", and UTF-8 (but not UTF-16) is safe to do byte-oriented searching on.