Ian Hickson wrote:
On Tue, 11 Apr 2006, Jonas Sicking wrote:
Ian Hickson wrote:
But I would add one more. Authors are stupid. We shouldn't provide them with
features whose only possible use is for them to shoot themselves in the
foot. In other words, I would phrase the question not as "which headers
should we restrict", but "which headers should we allow", and only allow
those that have valid use cases.
This sounds like what I suggested. But are there really any headers "whose only possible use is for them to shoot themselvs in the foot"?

Accept-Charset was the one that has been mentioned several times -- certainly unrestricting it (making it accept things that the UA won't know how to handle) doesn't seem very useful, since the UA will be unable to provide either a responseXML or responseText in that case.

Yeah, ideally Accept-Charset should be limited to charsets that the UA can deal with. Of course, requiring that the UA limits it to that might be a pain for implementors. Though probably not too bad.

However there are ways to use this header while not shooting yourself in the foot, so I think we shouldn't completely restrict it.

Though the usecase for Accept-Charset seems fairly weak. Why couldn't the author simply filter responseText?

As you could maybe tell, I don't feel strongly either way on this one.

/ Jonas

Reply via email to