> They don't fundamentally alter the nature of what a regex does, which is to recognize regular languages.
It's a bit subjective but captures are harder than recognition. Russ Cox has once noted [1] that the extraction has to be run as a separate step after the recognition and a fast DFA can't always be used for that, suggesting they are related but different problems.
Well, if you allow an arbitrary depth of capture-group nesting, then that may be so, but it seems beside the point here. It is not clear to me that this article makes any point about extraction that is relevant to this discussion.
It's a bit subjective but captures are harder than recognition. Russ Cox has once noted [1] that the extraction has to be run as a separate step after the recognition and a fast DFA can't always be used for that, suggesting they are related but different problems.
[1] https://swtch.com/~rsc/regexp/regexp3.html