Eric Brunner-Williams wrote: >>> no _technical_ reason that TLD labels should be all-alphabetic
>> FWIW, when you display internationalized domain names, and mixed RTL >> and LTR contexts (overall, in a label etc), you can get "interesting" >> results when characters that have not directionality (like numbers) >> are displayed adjacent to punctuation. >> See http://stupid.domain.name/node/681 for an example. > the example shows that the bidi algorithm used, correctly, if the > sequence of encoded values is "text", incorrectly, if the sequence of > encoded values is a dns label, render a directionality property of "." The only way to handle bidi with *PLAIN* text, which is what we use in UNIX command lines and DNS names, is to do it with finite state. Otherwise, you can't search, because search with push down automaton is prohibitively inefficient, which means nested directionality can not be supported. I warned so about 20 years ago. Then, bidi support with plain text is just simple. Characters do not have any bidi property and are displayed right to left in RTL lines and left to right in LRT lines (words are often spelled backward). However, as Unicode plain text includes control characters for nested bidi, and is not plain, its bidi support specification is totally confused and is unusable to handle real plain text. We can't move forward until we have character encoding with rational support for bidi. Masataka Ohta _______________________________________________ DNSOP mailing list [email protected] https://www.ietf.org/mailman/listinfo/dnsop
