On Apr 21, 2006, at 12:38 PM, Ian Hickson wrote:
On Fri, 21 Apr 2006, Mark Nottingham wrote:
There's a place for making sure you have a path from the current
implementations to the new standard, but this isn't it. Specifying
this
behaviour well isn't going to cost anything; some implementations
won't
be conformant for a little while, but fixing them won't break any
existing applications.
Besides which, each of those implementations does a different
thing; how
do you accurately represent that?
I wasn't saying the spec had to represent what happens _today_, that's
obviously impossible since the implementations differ. My point was
that
one of the implementors told you "we might never be able to do
this" and
yet it was still being considered.
Once an implementor says "can't happen", especially if they give quite
simple and clear reasons, there's no point going in that direction any
more. In this particular case, it means there's no point requiring
"put"
to be sent as "put" instead of "PUT" because an implementor has
informed
you that implementors use network libraries that are outside their
controls and these libraries sometimes convert "put" to "PUT".
Actually, I don't think the network library argument is a very good
one. If your network library is not capable enough, you should get a
new one or make your own. And in this case it was presented as purely
hypothetical, since Mozilla does in fact control their network
library. I don't buy that a project operated by the same
organization, hosted on the same server, and under a compatible open
source license can't be changed.
The argument that is convincing to me is that content currently uses
"get" and "post" and expects the method to be uppercased.
Regards,
Maciej