This is a very quick article I built to show a non-tech manager. Data is this good for display isn't the same as data that is good for user-input derived searching.

Text from non-en_UK source: Salvié’s Cucina, Rue de la français, 3522, Grünerløkka.
What a en_UK person would type: salvie's cucina, rue de la francias, 3522, grunerlokka.

This is why it may be a good idea to brutalise fields that get used as search indexes down to ASCII, so people can *quickly* type a nearly correct string, and get what they wanted. Also note minor performance gain, in most data systems, by expecting that people type in lower case; and storing the data in same context.

Some similar articles in worklog