On 17 Jun 2015, at 16:52, bert hubert wrote: > At least if the RFC does not specify it, we should pick something.
The over arching issue here is that there is no right answer regarding non ascii in URIs. A URI is a sequence of characters, but in HTTP the path must be ascii only, and can have %-encoded bytes -- without telling what charset is in use. Because of this, I would say one can in RDATA store a URI in any charset, but when the URI is used, it should have 8-bit bytes %-encoded. That said, libcurl do happily(?) send whatever bytes you give it over the HTTP connection, so %-encoding must happen before libcurl is used. Earlier versions of the draft suggested one should use UTF-8 for the URIs in RDATA, which imply among other things the path should be %-encoded, but the question then was whether it also should be normalized etc etc... And what if someone want non-UTF-8 in a URI? What we have to remember is that whoever knows the URI and creates the RR also run the web site (for example) that is to be accessed. Because of this, the only question has to do with the %-encoding. My PERSONAL preference would be to allow either %-encoded or non-%-encoded bytes in RDATA and that whoever consumes RDATA have to decide whether %-encoding should be applied or not. I would apply it if I where a programmer. :-) Because of this, Bert, I would store things in UTF-8, and hope for the best :-) Patrik
signature.asc
Description: OpenPGP digital signature
_______________________________________________ DNSOP mailing list DNSOP@ietf.org https://www.ietf.org/mailman/listinfo/dnsop