On 14 December 2016 at 21:53, Julian Reschke <julian.reschke@gmx.de> wrote: > On 2016-12-14 12:37, Martin Thomson wrote: > >> On 14 December 2016 at 21:51, Poul-Henning Kamp <phk@phk.freebsd.dk> >> wrote: >> >>> Well, UTF-8 would also go through HPACK, but by eye-ball it seems >>> that it would be more efficient. >>> >> >> If you have lots of ASCII still, you can probably Huffman encode, >> though if you have lots of non-ASCII, you need to watch out: a three >> octet UTF-8 encoded codepoint turns into (worst case) 82 bits. Best >> case is 58 bits (both of which are invalid, so maybe not). >> >> I can't remember, is there actually a good reason why we can't just >> start shoving UTF-8 in header fields? I mean, h2 is probably OK with >> this. >> > > Some APIs assume ISO-8859-1, so unexpected things might happen (of course > that's independent of the actual transport). > > Best regards, Julian > > ​Particularly since HTML4 taught us that "ISO-8859-1" means "Windows-1252", which actually has values for most of \x80-\x9F Cheers -- Matthew Kerwin http://matthew.kerwin.net.au/
Received on Wednesday, 14 December 2016 12:04:36 UTC