30 aug 2012 kl. 11:42 skrev Daniel-Constantin Mierla <[email protected]>:

> 
> On 8/30/12 11:22 AM, Iñaki Baz Castillo wrote:
>> 2012/8/30 Olle E. Johansson <[email protected]>:
>>>> AFAIK that's not "needed". When a request arrives to the proxy having
>>>> ;gr param in the RURI, the proxy extracts the gr value, "decodes" it
>>>> somehow (not need to have such a mapping in a DB) and gets the
>>>> associated binding, so just such a binding would be retrieved form
>>>> usrloc table when calling lookup().
>>> That means that one - knowing the algorithm - can reach all contacts 
>>> directly, regardless
>>> if they have a gruu or not. Or?
>> I expect that Kamailio generates a random key on startup for encoding
>> GRUU values and thus, it would be not possible to generate the same
>> GRUU value (for the same URI binding) without having such a key.
> 
> A random key at startup is making impossible to deal with gruu values in the 
> ongoing dialogs, because there was an encoding key before startup and another 
> one after, which is not able to decode previously encoded values.
> 
> Anyhow, the algorithm is known, because it is open source, but unique id 
> generation is taking in consideration server id, startup time, pid, a counter 
> or random (depending on option, for gruu is counter) and hash over AoR. All 
> together results in a very random value, where another encoding makes no much 
> sense.
> 
> This is for temp gruu, because for public gruu the rfc recommends usage of 
> instance value, which is set by client.
> 
>> 
>> 
>> 
>>>> 
>>>>> - If so, is that reachable information for the pua-regloc to publish the 
>>>>> gruu's?
>>>> Note that such a feature would require implementing RFC 5628:
>>>> 
>>>> http://tools.ietf.org/html/rfc5628
>>>> 
>>>> "Registration Event Package Extension for SIP
>>>>  Globally Routable User Agent URIs (GRUUs)"
>>> Yes. But to make it possible and to make it possible to restart kamailio 
>>> without loosing information, I think we have no other option but to store a 
>>> gruu flag in usrloc.
>> I see no problem at all in adding two new fields to the usrloc table:
>> gr_public, gr_temp.
> There is no need, the values for these fields are in the other fields. If the 
> client does not support gruu, the instance is not published, so it is stored.

Instance alone is not an indication of gruu support. 

/O

Attachment: smime.p7s
Description: S/MIME cryptographic signature

_______________________________________________
sr-dev mailing list
[email protected]
http://lists.sip-router.org/cgi-bin/mailman/listinfo/sr-dev

Reply via email to