Henri Sivonen

[Unsquashed view.](https://github.com/hsivonen/icu/commits/onetrie)

For optional dependencies, I think it's reasonable to let the dependency increase its MSRV before the entirety of ICU4X does. For non-optional util dependencies, I don't see the value of...

It appears that the ARMv7 specialization of dynamic shuffles was been bogus, and fixing the typos in 026b10ed273c9e3db424f2f48f219d79b4bbc0d0 revealed it.

Moreover, `` gets highlighted as if the slash was not part of the attribute value:  However, with a spec-compliant HTML parser, it is part of the attribute value due...

I'm so bad at GitHub notifications. Sorry. @SimonSapin, it's a bit unclear to me what "here" means. The [adoption agency algorithm](https://html.spec.whatwg.org/#adoption-agency-algorithm) has its outer loop count limited to 8 in...

> Could you explain the gist behind the Blink's limit, in that patch, My C++ foo isn't great. ReviewBoard went away, but [here is rebased version of the Gecko patch](https://bugzilla.mozilla.org/attachment.cgi?id=9031216&action=diff)....

The fix has finally landed in Gecko and is available for testing in Firefox Nightly.

That's really weird. I guess I will need to see in a debugger what's going on.

> The shortest path to have all three do the same thing would be for Gecko to change only not to replace no-break spaces when exporting plain text as plain...

I acknowledge that I have seen this, but I haven't had the time to evaluate this for performance impact.