Doesn't that make an hard life for the linter that now has to decide which template string represents a (complete) element and which is just text/whatever? I prefer the first solution because it's much simpler to parse