[Users] nat flag and branch routes
Daniel-Constantin Mierla
daniel at voice-system.ro
Wed Aug 31 12:34:50 CEST 2005
On 08/31/05 12:59, Bogdan-Andrei Iancu wrote:
> Hi,
>
> it works, but is not save since you can not be 100% that dst_uri
> presence is strictly related to NAT traversal. It's also used by RR
> module to force routing after loose_route; and by dispatcher for the
> same reasons.....
if you do not call other function that alter the r-uri except
lookup(location) as I said, I do not see why is not 100% sure that the
user is behind the nat. When loose_route() is used, lookup(location)
should not be used, I see no good reason. dispatcher is for load
balancing and it is usually in front of registrar. In this
circumstances, I would say that the situations to have many settings of
dst_uri is very less probable.
Daniel
>
>
> I see here two ways of approaching this issue:
> - to have per-branch flags also before transaction creation; will
> be a new param to append_branch (8 in total :-/), but this flags will
> not be accessible from script; only in branch route;
> - use something else than flags for NAT marking (something already
> present in all branch stages): nathelper, when builds the received URI
> (which will become dst_uri) will append a "nat=yes" parameter; this
> parameter will be easyly identify in branch route and NAT traversal
> may be activated....
>
> any comments or new options are welcomed.......
>
> regards,
> bogdan
>
> Daniel-Constantin Mierla wrote:
>
>> I would say yes, if you do not call other functions that alter the
>> r-uri/dst_uri, except lookup("location").
>>
>> Daniel
>>
>> On 08/30/05 19:43, Richard Z wrote:
>>
>>> Just a thought... is it possible to ingore the nat flag and just
>>> rely on the existence of dst_uri to indicate a NATed UA?
>>>
>>> On 8/29/05, *Klaus Darilion* < klaus.mailinglists at pernau.at
>>> <mailto:klaus.mailinglists at pernau.at>> wrote:
>>>
>>> Ho Bodgan!
>>>
>>> To use branch routes for branch-only NAT traversal also the
>>> nathelper
>>> and mediaproxy functions must be adopted to work in branch routes.
>>>
>>> regards
>>> klaus
>>>
>>> Bogdan-Andrei Iancu wrote:
>>> > Hi,
>>> >
>>> > indeed, prior branch_route, there is only one set of flags
>>> shared by all
>>> > branched - that's still unchanged.
>>> >
>>> > regards,
>>> > bogdan
>>> >
>>
>>
>
>
More information about the sr-users
mailing list