Re: -encryption draft -01

On 2014-12-16 22:31, Roy T. Fielding wrote:
> On Dec 16, 2014, at 9:56 AM, Julian Reschke wrote:
>> On 2014-12-16 18:46, Martin Thomson wrote:
>>> Feedback, structured or not, is always welcome.
>>>
>>> I didn't realize just how riddled this was with little bugs.
>>>
>>>>     A client can also explicitly probe for an alternative service
>>>>     advertisement by sending a request that bears little or no sensitive
>>>>     information, such as one with the OPTIONS method.  Likewise, clients
>>>>     with existing alternative services information could make such a
>>>>     request before they expire, in order minimize the delays that might
>>>>     be incurred.
>>>>
>>>> Q: How is OPTIONS better than HEAD?
>>>
>>> I believe that either is fine.  This is a f'rexample only.  I think
>>> that we had a discussion where (and I'm going to rely on bad memory)
>>> Roy suggested OPTIONS over HEAD.  OPTIONS * allows a client to learn
>>> things without perhaps revealing what resource it might be interested
>>> in.
>>
>> I don't think that doing a request for "/" will reveal more than a request for "*".
>>
>> Anyway, I'd like to avoid anything that might get people to think that "OPTIONS *" is anything but a bad design mistake. In particular I don't want any new specs to rely on it, when it's known to be hard to implement (as the asterisk form doesn't map to a proper URI).
>
> It isn't hard to implement.  Please stop with the FUD.  OPTIONS is
> preferred because it does not cause the server to do a retrieval
> (HEAD is usually implemented as an internal retrieval that gets
> truncated after the header fields, so it has almost as much cost on
> the backend as a GET).

"OPTIONS uri" is fine, "OPTIONS *" is a problem...

Best regards, Julian

Received on Tuesday, 16 December 2014 21:36:48 UTC