On Sunday, June 1, 2014 2:01:09 PM UTC+5:30, Chris Angelico wrote: > On Sun, Jun 1, 2014 at 5:58 PM, Marko Rauhamaa wrote: > > As a Finnish-speaker, I hope that patch doesn't become default behavior. > > Too many times, we have been victimized by the German conventions. A > > Finnish-speaker would much rather see > > Järvenpää => Jarvenpaa > > Öllölä => Ollola > > Kärkkäinen => Karkkainen > > than > > Järvenpää => Jaervenpaeae > > Öllölä => Oelloelae > > Kärkkäinen => Kaerkkaeinen
> It's even worse than that. The rules for ASCIIfying adorned characters > vary according to context - Müller and Mueller are different names, > and in many contexts should sort and compare differently, and I > remember reading somewhere that there's a context in which it's more > useful to decompose ü to u rather than ue. There is no "safe" lossy > transformation that can be done to any language's words, and this is > no exception. ASCIIfication has to be accepted as flawed; this issue > (an inability to handle non-ASCII labels) is similar to a lot of blog > URLs - > http://rosuav.blogspot.com/2013/08/20th-international-g-festival-awards.html > is talking about the "International G&S Festival" awards, but the URL > drops the "&S" part. (If you absolutely have to transmit something > losslessly in pure ASCII, you need a scheme like Punycode, which is a > lot less clean and readable than a decomposition scheme.) > Of course, the better solution is to permit the full Unicode alphabet > in identifiers... Yes that is the real point. Changing the current behavior which maps [ö,ä…] → [o,a…] to a new behavior that maps it to [oe,ae…], then arguing that this should/should not become default is the wrong battle. The more useful line is: Why have this conversion at all? Until hardly 3 years ago html authors wrote non-ASCII as chars as html entities. Now the current standard practice is directly to write the character and make sure the page is explicitly utf-8. Its only a question of time before this becomes standard practice in all domains -- https://mail.python.org/mailman/listinfo/python-list
