> >Looking out the rear-view mirror, this is true. Looking out the > >front-windshield, this is not true. > > no doubt your base of experience permits you to speak with such > certitude. those of us who have worked on Internet technology for less > time, say only 20 or 30 years, find that such predictions are usually > wrong. sometimes right, but usually wrong.
My "base of experience" gives me an ability to follow current activities well enough to see that there is more active development for UTF-8 than any other charset. Do you dispute this? I can infer nothing else from your comments above. > The physics of upgrading a large, infrastructure-critical service > dictate a text-based encoding scheme. Debating this point requires more context. Certainly IP, BGP and many of the other "large, infrastructure-critical services" don't rely on text-based encodings. DNS itself is 8-bit clean, and only uses a text-based encoding for payload data, and that's only to support the legacy HOSTS.TXT requirements. What particular aspect of UTF-8 isn't "text-based" anyway? DNS itself doesn't even use a text-based encoding (supporting 0x00-0x1F and 0x80-0xFF natively) by the yardstick you seem to be applying. > It simply has no effect on significant system performance issues, other > than transparency to the existing system... if an ACE is used. Transparency to users, admins and developers is also important. Think about ACE in terms of everyday usage scenarios, particularly in markets and development environments where people are trying to internationalize their networks and services. The educational and implementation burden of an ACE-only approach is massive. Worse, it is perpetually massive. Allowing the users and developers to lower their implementation costs in a non-disruptive form should be considered as our duty. Preventing it without reasonable cause is pointless. > UTF-8, on the other hand, will nicely break the existing system. Dan J. is not arguing for UTF-8 *ONLY*. That's a non-starter. The UDNS proposal allows for a dual-mode structure, one of which is ACE for legacy-systems compatibility, the other of which is a UTF-8 for future application support. They are COMPLEMENTARY. -- Eric A. Hall http://www.ehsco.com/ Internet Core Protocols http://www.oreilly.com/catalog/coreprot/
