conrad.irwin wrote:

You can already include un-normalised text using HTML entities, see [[wikt:Appendix:Unicode normalization]], or as you did with manual URL encoding.

I reckon most people who know why such things are broken will be able to work out the escapes, though a utility to help them wouldn't be bad. For any given text, it requires detailed knowledge of unicode and the context to identify the few cases where normalization is destructive.

Unicode normalization is done on input, so it's hard to imagine the tag method you propose working safely and efficiently.

ayg wrote:

Should we avoid Unicode normalization in URLs altogether? It seems as though it's likely to cause this sort of problem.