Re: p1: Via and gateways

[Mark Nottingham <[e-mail]>]

OK, it seems like we agree we can relax the MUST generate Via on responses from gateways, provided that they don't removing any existing Via header. 

I'll mark this for -23.


On 24/04/2013, at 4:38 PM, Willy Tarreau <[e-mail]> wrote:

> On Wed, Apr 24, 2013 at 01:27:45PM +1000, Mark Nottingham wrote:
>> 
>> On 23/04/2013, at 4:43 PM, Willy Tarreau <[e-mail]> wrote:
>> 
>>> Then I would propose this addition :
>>> 
>>>  Multiple Via field values represent each proxy or gateway that has
>>>  forwarded the message.  Each recipient MUST append its information
>>>  such that the end result is ordered according to the sequence of
>>> -  forwarding applications.
>>> +  forwarding applications. A gateway MAY simply relay any existing Via
>>> +  header field if it does not change the HTTP version, but it MUST NOT
>>> +  remove it.
>> 
>> This is already covered further down:
>> 
>>> A proxy or gateway may combine an ordered subsequence of Via header field
>>> entries into a single such entry if the entries have identical
>>> received-protocol values.
> 
> I don't read it exactly the same way, but probably it achieves the same in
> the end.
> 
>> The question I was raising was specific: can we relax the requirement for a
>> gateway to add Via to responses if there isn't already one present?
> 
> I think yes, we can relax it, as if there was none, then it means the
> gateway has not changed the connection's behaviour and the next hop just
> assumes the connection comes from the client. Again, in my understanding,
> Via is used to understand what transformation was operated on the path
> and what capabilities are offered. If the gateway does not change anything,
> Via offers no value except detecting loops. And since most gateways simply
> forward to the configured next hop, the risk of loop is not caused by
> external environment (eg: DNS) but by the configuration so it's not a
> problem.
> 
> But if we do so, we must specify that a gateway MUST NOT remove the
> Via header field in any case, otherwise it will break the loop detection
> mechanism of some other components.
> 
> Regards,
> Willy
> 

--
Mark Nottingham   http://www.mnot.net/