Misencodings as aliases

Tags: #<Tag:0x00007f00c2cee360>

In our alias documentation we have a possible use of aliases:

  1. Misencodings :

Names entered in FreeDB using non-UTF-8 encodings; these are somewhat like typos, but in non-Unicode locales, these may in fact be more accurate than an automatic conversion from UTF-8.
e.g. ©PªN­Û­ = Jay Chou

There’s a ticket (MBS-5193) about how some control characters in these, such as reverse line feed, are removed from the aliases, breaking them.

The comments on the ticket suggest that the right thing to do is just to drop this kind of misencoding as one of the things we do with aliases, since the 90s are over. IMO that might very well be the right decision, but:

  1. Do some people disagree that it makes sense to stop doing this?
  2. Is it worth it to actually remove this kind of aliases when they exist, or should they just stay there?

There seems some benefit in having misencodings in Aliases.

What are the costs/downsides?