So what's being discussed here is basically reading/writing raw data directly to the in/out streams of the HTTP request. My PHP is rusty, at best, these days, but this would involve processing a request by directly reading the input using something like
this instead of accessing things via the POST or GET hashes. And clients would be expected to just dump the raw JSON encoding to the input stream instead of a url encoded html form of the json encoding.
I agree it'd be easier for the client authors if this were done, but not so much easier, imho, that it's worth deprecating any test drivers that may be written using HTML forms, etc. I only say that because, especially for a language like C#, I assume you're using some kind of library to construct the HTTP requests? Encoding the data before sending it along the wire can't be more than a simple method call (and that's only if it's not done "automagically" by the lib you're using). For me, in Java, I'm using the Apache HttpClient lib for all HTTP traffic and the encoding/decoding is done for me without any thought on my end.
All of my FVT tests for the Java client are written in pure Java and don't use any HTML forms, etc. so I'm not worried too much either way. If the switch were made it'd break all the Java client code and so I'd have to fix it so I'd just fix my test cases as the same time. But if someone has web pages for manually testing the service, they may need some serious rework, depending on how they were currently implemented.