Jack Wrenn

Thank you so much for your review, @Havvy! If it's alright, I'm going to defer removing the feature flags until this gets a little closer to being merged; otherwise, I...

> @jswrenn Are you still working on this? If not, I would like to work on it. I'm not! Go for it! (Sorry for the delayed reply; I thought I...

@hawkw, whoops, I definitely meant to write: > Should it take `&self`, or `&mut self`? An argument in favor of the method consuming `&mut self` is that it'd mirror `Layer::on_layer`,...

My feeling is that it's probably uncommon to have many complex subscribers initialized and installed over a program's lifetime, so the risk is low. In the near-term, it's sufficient to...

> I'm also open to investigating moving the thread local collector features out of tracing-core and into a higher-level crate. For this, does the below sound like a viable gameplan:...

I've updated the issue to include a [flamegraph](https://jswrenn.com/misc/pywb_573-profile.svg) of `wayback` handling a single request, initiated by `curl`. A substantial amount of time appears to be spent searching the index.

Multiple index files: ``` indexes/ ├── [596M] autoindex.cdxj ├── [118M] autoindex.cdxj.tmp.20200629011258909273 ├── [ 0] autoindex.cdxj.tmp.20200629152048161219 ├── [ 0] autoindex.cdxj.tmp.20200630020638761945 └── [611M] index.cdxj ```

I deleted the existing indices and re-indexed the collection, but there was no improvement.

> If there's any way you can share the example privately, I can try to debug further.. Would sharing the index be sufficient? (There's a bunch of FERPA-restricted and NDA-restricted...

Thanks!!! I'll send that email momentarily.