The new draft is available at:

If you gave feedback you were expecting

to be addressed, please double check that it was.

In the background I've tried to tone down the

importance of regexps while expanding very

briefly on their history.  Specifically, their

raison d'être is that they can be matched

efficiently.

I've removed the string argument to the

rx-match-* functions and renamed them

regexp-match-* as discussed.  The primary

matching functions are now:

  regexp-search

  regexp-matches

  regexp-matches?

I removed the backtracking-specific SREs

if, atomic and commit, because they are

specific to the particular backtracking algorithm

they are used with and thus too difficult to

specify.

The remaining extended features I've split

into regexp-non-greedy and regexp-look-around.

There is still debate whether the features should

be provided at all.

I've given a more concrete definition of

grapheme clusters, and noted that case

insensitive matching does not have to take

into account characters with a non 1:1 mapping

like ß.  I still need to expand more on Unicode

word boundaries.

Mutating a string that was matched now invalidates

any regexp-match results, and likewise mutating

an SRE invalidates any regexps compiled from it.

-- 

Alex