--On Thursday, June 8, 2017 09:58 -0700 Randall Gellens <rg+ietf@xxxxxxxxxxxxxxxxx> wrote: > I personally find the effects more an amusing artifact that an > actual hindrance, but that's just me. However, when I'm > traveling and not part of an IETF, I find that using a hotel > provided IP address without a full VPN causes its own set of > difficulties with web sites assuming the language and > character set of the country I'm in. In some cases I'm unable > to use a Capcha since it's in a script I can't read, and I'm > unable to select a language or country on the web site for the > same reason. While I agree, I think this may be a different problem and one that we (and/or W3C) could fix if we were motivated. A variation on the "incentives" theme is useful here too -- very few actors have incentives to condition information being provided to you based on bad choices of language or locale, even if there are complex case-by-case decisions (or sheer laziness) about whether the costs of getting things right in a larger number of cases are worth it. We've got machinery for specifying preferences in cultural or locale information to web browsers (and the SLIM WG has been working on somewhat parallel arrangements for other kinds of real time communications). Are those options right? Do they provide enough preference information that you (or a typical user, or a heuristic) can figure out whether to use the information they think you've given a browser (or failed to provide), rather than what the hotel, or the hotel network, or even the choice of computer if you are using one you don't own) are telling them? Those are actually not easy questions in terms of designing the right sorts of tools and indicators. Pursuing your example, suppose you are in Lower Slobbovia, don't speak or read Lower Slobbovian, but want to find a nearby pizza restaurant. Probably you want to have the system communicate with you in English, but you don't want the locale information to reflect, e.g., California, because knowing where the nearest pizza restaurant would be if you were in California would not be of much interest. That, in turn, leads to another economic issue: if the local, Lower Slobbovian, restaurant operators federation has concluded that it isn't worth the costs of supplying information in English, then you don't even have a localization issue any more, you simply have a choice between asking in Lower Slobbovian (or some other language they want to support) or not asking in any way that yields answers. On can respond to that example by asking questions about automated translation, etc., but I'd just make the example more complicated. It seems to me that set of issues has little to do with how the servers that interact with the hotel network guess at language or location and, in particular, whether they use the IP or MAC address to get there. But it does seems to me that it suggests we should review whether our various means for explicitly telling systems about locale information do what is wanted and fix them if they don't. Certainly, such fixes should recognize the right of those who don't want accurate location information under any circumstances to starve in the dark, but it is not clear to me that they should be our primary audience. john