Re: [tcpm] Exceeding value in MSS option?

[Frode Kileng <[e-mail]>]

On 23/10/2020 17:11, Joseph Touch wrote:
>> On Oct 23, 2020, at 1:37 AM, Frode Kileng <[e-mail] 
>> <mailto:[e-mail]>> wrote:
>>
>> Although maybe not  a "general problem",  a study of MSS middlebox 
>> manipluation in  23 European and Asian mobile networks did reveal 
>> some issues:
>>
>> - 20 of 23 networks always inserts MSS.
>
> That IS a general problem, FWIW.
>
> Besides being published, it should also be reported to the operators 
> of those networks as violating TCP specs.

Do you mean as in "layering violation" or is it explicitly stated in an 
RFC that MSS clamping is not allowed.

Anyway, since stopping this practice is likely to cause extensive 
fragmentation of TCP segments, mobile networks will continue this 
despite "violating TCP specs". At least until the networks fully support 
jumbo frames.

(The motivation for MSS clamping is the tunneling overhead in the mobile 
core of the networks. Within the core, the MSS is clamped to take into 
consideration the size of the  GTP-U header and potential additional 
ipsec headers. The GTP-U header is variable size and is at least 40 
bytes for IPv4 and 60 for IPv6. In addition it may include vendor or 
mobile operator specific extension headers. I do not now why networks 
don't reduce the MTU size instead. My understanding of the 3GPP specs is 
that this is suggested but could not find out how this is provisioned to 
the terminals for IPv4. I've only seen one mobile network that 
provisioned the terminals with a MTU lower than 1500 bytes. But their 
MTU size of 1464 for for IPv4 and IPv6 did not seem reflect the GTP-U 
tunnel.)


Regards

Frode Kileng