Sergei Lebedev

JFYI there's another [blog post](http://julesjacobs.github.io/2015/06/17/disqus-levenshtein-simple-and-fast.html) with Python code for building and using the automata. > I may look into implementing a fuzzy (branching) search in marisa-trie in the near future....

I'm not sure if this is possible using the [public API](https://marisa-trie.googlecode.com/svn/trunk/docs/readme.en.html) of `marisa-trie` the C++ library. Need to check and then maybe forward the issue upstream.

@jorenham I've just started [a thread](https://mail.python.org/archives/list/[email protected]/thread/4T2JALYTXODIFQ5HYBM7RTYK2UU7AG4Y/) about exactly this on typing-sig. Would would be the best way to provide feedback on the PEP draft?

@fylux The proposed syntax uses unary +/- operators, so it shouldn't interfere much with type level arithmetic, which I assume would mostly use the binary ones. `A ++ B` looks...

> I agree that `ParamSpec` should not allow a bound type, type constraints, or variance parameters. Is it too late to amend the PEP and remove the mention of these...

@hhugo bringing this into the main APIs would probably break a lot of code, would you like a separate labelled module a la `StdLabels`?

@erictraut can you elaborate on why `Not` will not compose with other features? Are you implying that there is no sensible way to incorporate `Not` into subtyping? `Not` type arise...

Thanks, Eric! These are all great points. I can suggest natural answers to most of your questions, but I agree `Not` is tricky and will take a lot of work...

It might be tricky to support `scatter-add` in its entirety in Triton, but maybe we can support a handful of useful special-cases.

Thanks, Yuxuan!