Why does google's home page declare conflicting character encodings?

#discuss #webdev #webperf #performance

Edit: Still don't know why. But I did find out it only happens if you've got a weird or missing user-agent string in your request headers.

I've been doing some research on declaring character encodings.

Specifically, do you really need the <meta charset="UTF-8"> tag?

You must declare a character encoding, but by default most servers include this in the http headers and that's actually better than using a <meta> tag — the earlier it's declared the sooner the page can render.

A micro-optimisation really.

On top of that, for HTML5 utf-8 is the only valid character encoding. So <!doctype html> is implicitly declaring the character encoding too.

<meta charset="UTF-8"> is considered sacred. So before I started telling people it's a useless 22 bytes. I thought I'd see what google do.

In the google homepage <head> tags they have:

<meta content="text/html; charset=UTF-8" http-equiv="Content-Type">

But then in the http headers it's:

Content-Type: text/html; charset=ISO-8859-1

What's going on here?

Here's my guesses:

Maybe it's a backwards compatibility thing. Perhaps browsers that don't understand the <meta> tag also don't understand utf-8?
Maybe it's a performance optimization. Perhaps it's faster to parse the very first part of the document in ISO-8859-1 then switch to utf-8 for the rest.

What do you think? What does google know that we don't (besides literally everything)?