On Sun, Apr 09, 2017 at 04:30:36PM +0000, Stuart Stuple wrote:
My perspective is that Universal Acceptance means that the user gets to decide how these are consistent or diverges (rather than technological limitations).
If that is your goal, I hope you realise you're going to be disappointed. Natural language processing is enormously complex and subject to a great deal of contextual error-correction and retransmission and so on. The basic insight of the Davidson "prior theory/passing theory" interpretation of how language works is that nodding of heads, "uh huh", "ok?", "get me?" and all that is not extraneous to language use, but is a deep and abiding part of it. The reason our written language gets more formal -- as we go from text messages to personal letters to emails on distribution lists to job application letters to proposals to contracts to protocols and laws to constitutions -- is because the context is gradually stripped as we ascend that staircase. It's easy for your friend to send you a ? in the even t she doesn't understand your text messasge. We have a hard time doing that to the authors of early RFCs, never mind to the authors of the US Constitution. Identifiers in computer systems are even worse than constitutions, because they have practically no context at all. This is _especially_ bad in the case of Internet identifiers, which are mostly free of almost all context, including the linguistic context of the different points involved in the network transaction. (I say "different points" because of course they can be multiple. In an email address, for instance, there is the language of the sender, the language of the receiver, the languages supported by the mail operator, and the linguistic context rules supported by each domain name administrator at each domain on the domain name path -- a context that need not be consistent down the tree.) There may be some future state in which networks have the necessary metadata to perform the right kind of linguistic negotiation along the path. At least until computers are sufficiently good at natural language processing that no human speaker of any language will detect them as being computers, that state will remain in the future. I think we can achieve some more modest goals in the meantime. Best regards, A -- Andrew Sullivan ajs@anvilwalrusden.com