User tracking has been discussed a lot during the development of the QUIC 
protocol. That is not say the discussion is no longer relevant, but it has not 
been ignored.

For servers, it is necessary to track users across migrations, because you need 
to maintain connection state and to maintain the IP address of where to send 
data. One could add meta-criteria outside the protocol such as the QUIC 
application must not extract IP data or other user identifiable data from the 
stack, but that is not enforceable within the protocol, and it is also not 
always desirable. For example when having to whitelist users.

On the other hand it is important to not allow servers to track users across 
different connections. This is complicated because of 0-RTT state being 
maintained. I think perhaps more work could be done in this area for future 
versions. On the other hand there is wording in the protocol to prevent 
external observers from linking different connections to the same user, 
although it may still be possible through traffic analysis which is also 
mentioned as a limitation, if I remember correctly.

Mikkel


> On 7 Jun 2021, at 15.04, Stephane Bortzmeyer <[email protected]> wrote:
> 
> On Mon, Jun 07, 2021 at 01:52:19PM +0100,
> Lucas Pardue <[email protected]> wrote 
> a message of 43 lines which said:
> 
>> As Robin says, to survive such client IP changes would require QUIC
>> connection migration. RFC 9000 Section 9.5 [1] deals with the privacy
>> implications of migration.
> 
> This section is completely silent about tracking BY THE SERVER. It
> only considers a passive observer, which is not the only threat.
> 

Reply via email to