On 17-Oct-22 16:57, Daniel Stenberg via curl-library wrote:
On Mon, 17 Oct 2022, Dmitry Karpov via curl-library wrote:

I'm all for adding an option to add the host-only behaviour as an option, but not to make it the default.

Yes, I also think that this is the right way to do it.

I hear you. Thanks all for the feedback.

If we're going forwward with this, the new behavior should be a new option. But, as this is a rarely used option already to begin within I'm not convinced I should push for adding this new option until someone actually expresses a desire to have it...

It can be a new value for the current option, you don't need a new open.  Just don't change the behavior of the current documented values (0 & 1).  The API & ABI need to be stable.  The doc says that a value of 1 "automatically sets the Referer header ... "to the FULL URL when it follows a redirect".

E.g. make the "strip path" value 2.   Then existing callers will get the full path that they currently do, and and new code can decide what works for them.  You could define an enum with 2 as "CURL_OPTVAL_REFERRER_STRIPPATH" and 1 as "CURL_OPTVAL_REFERRER_KEEPPATH".  (and zero as "OFF")

I've no problem with the doc explaining the privacy issue and even suggesting that STRIPPATH is a better choice in many (most?) cases.  Just so the API and ABI are stable.

Yes the browsers often break things - as Dan confirmed.  That doesn't make it right.  And they have the advantage of a human being able to change a setting; curl often runs in scripts or applications that aren't as smart.

Onwards.


Attachment: OpenPGP_signature
Description: OpenPGP digital signature

-- 
Unsubscribe: https://lists.haxx.se/listinfo/curl-library
Etiquette:   https://curl.se/mail/etiquette.html

Reply via email to