On Tue, Aug 16, 2011 at 8:10 PM, Robert Newson <rnew...@apache.org> wrote: > "The only requirement is: if I ask an HTML, give me an HTML, not a JSON." > > If only it were that simple. How about, as in the case for IE, both > HTML and JSON are exactly equally acceptable? 1.0.x would send HTML, > 1.1.x will send JSON. Both are wrong depending on what you wanted. > Empasse.
I am starting to think this question is similar to the Futon test suite question, and the question about whether CouchDB should know about nonstandard proxies. The right answer, or at least the only workable one, is to keep doing what CouchDB does: be optimistic, and assume the future will be better than the present. Specifically, take the web seriously. Follow the standards and best practices (the REST model, and ETags come to mind). Because future CouchDB might work one of two ways: 1. It doesn't always work on every browser, or through every proxy, but it is consistent. You can make assumptions about it. 2. It doesn't always work on every browser, or through every proxy, and it is also inconsistent. In case #1, at least application tooling can, in principle, discover what's going on and do the needful. For example, add ?rnd=$RANDOM to every GET. In other words, follow the standard. If the community needs an IE workaround, we can discuss that then. -- Iris Couch
