On Wed, Aug 27, 2008 at 2:49 PM, Pramod Sharma <pramodsharma08_at_gmail.com> wrote:
>>>From: Phil Blundell <pb_at_reciva.com>
>>Date: Mon, 11 Aug 2008 14:51:34 +0100
>
>>you need to URL-escape the string (i.e. convert % to %25) after adding
>>the scope. So, try:
>>curl_easy_setopt(curl,CURLOPT_URL , "http://[fe80::212:3fff:fe55:8470%252]:80/SiteStat.xml");
>>You don't need to set CURLOPT_SCOPE if you include a scope id in your URL.
>>p.
>
> Hi ,
> This is just to confirm my understanding are correct.
> 1. This CURLOPT_ADDRESS_SCOPE is generally the interface id for the
> Link Local address.
> 2. Now if proxy comes in between , he will have to take care of that
> Link Local address somehow like attaching the interface id with
> address.
> I don't think curl can take care of that thing at all as that proxy
> should attach his interface id with link local address to connect
> otherwise connection will fail. ?
>
> Why I am saying all this is because of this
> ./curl --verbose -6 -g http://[fe80::212:3fff:fe55:8470%252]:80/ttttt
> --proxy [2001:db8:0:1:203:ffff:fee1:2a22]:3128
> is failing as proxy is failed to connect to fe80::212:3fff:fe55:8470
> , here "2" is the interface id (of the local machine ) passed to curl
> and if no proxy then things will go perfectly as curl can establish a
> connection using this interface , but if proxy in between he will have
> to take care of that. ?
>
> -Pramod
>

>>Yes, I believe your understanding is basically correct.

>>I'm not convinced that it makes much sense to have a proxy involved in
>>connections to link-local addresses, particularly not a
>>globally-addressed proxy that might be on a different link to the host
>>making the original request. I suspect that most UAs would be better
>>off bypassing the proxy when the URL contains a link local address.

>>p.

Thanks Phil. Yes you are right about that it doesn't makes sense
putting non-global address through proxy. Only thing is I am trying to
support upto whatever we can ...rest leave it to the world..
Received on 2008-08-28