Comment 1:

Note that this feature was requested on the ML by someone who was showed an easy
solution to the case he presented. He only said his case is "more complex" but never
shown a particular "complex" case which current regexp package cannot solve. As the guy
even seemingly never accepted the fact that package regexp does not backtrace, I tend to
think that his understanding of REs is not sound and his non presented actual use case
probably _is_ covered by the current regexp package. Also I think the implementation
would be problematic - excluding certain states from a state set lazily is IMO not very
compatible with what (executing) a NFA is, and it sounds like backtracking.

Comment 2 by amit.lavon@mail.huji.ac.il:

Had the commenter above read the discussion thoroughly, he would have understood that
the requester had never asked for a solution for the simple example he presented. He
only asked why there are no possessive quantifiers in go - which no one answered.
Yet people who don't understand what possessive quantifiers are jumped up and tried to
prove the requester wrong without reading the question, as did the guy above, making
arrogant comments about the requester's background.
Please relate to this feature request as "it would be great to have ..." instead of
"find me a solution for ...", same way we can parse arguments without the flag package
but it still exists. So possessive quantifiers can be very helpful, even if you
specifically can't think of a uses for them. Because there are.

Comment 3:

Had the commenter above read the discussion thoroughly, he would have noted that he has
been told 3 or 4, maybe even 5 times that package regexp does not backtrack. Yet he
insisted that he "proved" the backtracking. As that is absurd, any conclusions derived
from the false fact are invalid. Truth is not arrogant.
You are however right that you've never asked for a solution to a problem you have. And
that's the point: You might in that case have got a satisfactory solution. Who knows...
Conclusion: You're asking for a feature which you claim, without any proof presented,
solves a problem which you didn't even show to anyone. Notice that I'm not saying the
feature request should be accepted or rejected. I'm just pointing out that there's no
evidence the change provides anything worth the work needed to possibly implement it.
Also, it seems to be the first regexp feature request for this very functionality in the
sixth year Go is public. Additionally, I still have a feeling the change may be equal to
the introduction of backtracking (far from sure I'm).
""""
possessive quantifiers can be very helpful, even if you specifically can't think of a
uses for them. Because there are.
""""
The burden of proof is obviously at your side. Show a concrete problem which it solves
and where the current regexp doesn't help and you can easily win your case by rational
argument. Hand waving, seen so far, is going to convince few, if anyone.
If you're still interested in this kind of conversation, let's please return it to the
ML. This place is not appropriate.

rsc wrote in a comment in CL 67590 that

the regexp syntax is essentially fixed at this point, which is to say immutable.

So I think this can be closed. Closing.