Xiaocheng Hu

Can we leave it unsupported? I think it's totally fine to proudly leave a deprecated feature unsupported. Especially when its [usage](https://chromestatus.com/metrics/feature/timeline/popularity/3009) is very low (0.2% as of today), which means...

> This is one of the trivial issues I opened from time to time and I wish to encourage contributors to be able to get some tasks. Thanks for the...

Discussed with @stevennovaryo offline based on his design doc https://hackmd.io/@ChaKweTiau/H1KmdqqMxg Creating a cache of all parsed UA sheets seems good and quite approachable. However, the best approach seems to be...

@stevennovaryo That sounds reasonable to me! As we know, the performance of shadow DOM can be improved as a whole in the future. Ensuring the correct behavior seems more important...

The last run was all green. Is it ready for review now?

Per spec, we should use UTF16 code unit index instead of code point index. How about a more holistic fix: replace all strings with a UTF16 string class?

Are there any high priority bugs that require this ad hoc fix? > I'm also not sure if there's even an observable difference between utf8 and utf16 in this case,...

All right, I'm fine with this PR as is. I'm not a fan of adding conversions here, but I'm even less a fan of making a temporary solution perfect...

Is there a dedicated issue / discussion thread for the changing the backing impl of strings? Because it's not just limited to `textinput`. Wherever we need to expose a string...

> If it isn't then we have to do this transition incrementally, and I figure the text input is a good place to start since it is fairly isolated from...