I'd be interested to see the logs to see if anyone is selecting a build that differs from their identified user-agent. If it is a problem, it would be easy enough to bundle a small executable that initiates the correct download. Or is that silly. Stephen On Sunday, August 14, 2011, Stephen Bloch <<a href="mailto:sbloch@adelphi.edu">sbloch@adelphi.edu</a>> wrote: > > On Aug 13, 2011, at 8:52 PM, Eli Barzilay wrote: > > "They [...] were presented 5 different options for downloading > DrRacket for Mac". > > Yes, but was one of them clearly distinguished from the rest, to > indicate "this is probably what you want"? > > Yes, by being the first -- which is the same as things are now -- > which is why I bothered with trying to make it less visible. > > If a user KNOWS that the web server has made a probably-correct guess at which version the user needs, that's great. > But by definition, we're concerned with users who have little or no technical background, and probably very little technical confidence. Such a user is left staring at a long list of slightly-different version names, with no basis for choosing one over the others but a very real fear that choosing the wrong version will cause the computer to explode. The probably-correct answer has to be SEPARATE from the rest, to give the user confidence. > A minimal solution would be to use the current format, but add the text "We've made our best guess as to which version you need. If you're sure this guess is wrong, feel free to select a different one from the menu. If you pick the wrong one, the worst thing that can happen is [FILL THIS IN]." > > Stephen Bloch > <a href="mailto:sbloch@adelphi.edu">sbloch@adelphi.edu</a> > > > -- -- Stephen De Gabrielle <a href="mailto:stephen.degabrielle@acm.org">stephen.degabrielle@acm.org</a> Telephone +44 (0)20 85670911 Mobile +44 (0)79 85189045 <a href="http://www.degabrielle.name/stephen">http://www.degabrielle.name/stephen</a>