Re: Feedback on the ping="" attribute (ISSUE-1)

Henri Sivonen wrote:
> On Nov 3, 2007, at 01:10, Julian Reschke wrote:
> 
> [Hixie already reiterated that this is about idempotence--not "safety", 
> so I'm not replying to the safety points.]
> 
>>> On Sun, 28 Oct 2007, Henri Sivonen wrote:
>>>> That might work and could be a tad safer. It isn't in any way 
>>>> theoretically pure from the RFC 2616 point of view, though, to make 
>>>> HEAD and GET have different semantics beyond the response body 
>>>> presence.
>>> Indeed.
>>
>> Yes, and nobody suggested that.
> 
> It should be obvious that if browsers consistently used HEAD for pings, 
> servers would have a *very* strong incentive to give HEAD and GET 
> different semantics beyond the presence of the response body. Thus, the 
> net effect of browsers using HEAD for pings would lead to a semantical 
> divergence of HEAD and GET even if it weren't explicitly suggested or 
> were explicitly suggested against.
> 
> Consider this from the point of view of an ad server. If user-operated 
> browsers are known to use HEAD, you can be pretty sure that a GET is not 
> an effect of a user following a link in a ping-aware browser. Thus, all 
> GETs are noise. Ad servers care more about quality impression data than 
> about RFC 2616, so it would make perfect sense to ignore GETs but not 
> HEADs.
> 
> The larger effect of this would be that the trustworthiness of HEAD as a 
> reliable way of obtaining metadata about GETtable resources would be 
> diluted.

I did *not* recommend to use HEAD instead of GET. HEAD and GET have the 
same semantics, expect for the response body not being present for HEAD.

A server that implements a ping target resource has the choice what 
representation to return for that URI. The cheapest and sanest thing to 
do would be to return an empty body, thus making GET and HEAD behave 
*exactly* the same.

Finally, my understanding was that the ping URI does *not* point to the 
ad server (which sees the GET on the href target + the Referer header), 
but for the convenience of the site *containing* the hyperlink.

Best regards, Julian

Received on Saturday, 3 November 2007 09:50:56 UTC