Mirimir wrote:
On 02/03/2020 02:17 PM, s7r wrote:
<SNIP>
In the current form of this proposal, it looks kind of optional ("We propose this optional change, to improve..."). I propose removing the line which contains "this optional change" and changing the following:
In practice, each operating system has a different way of detecting IPv6 address privacy extensions. And some operating systems may not tell applications if a particular address is using privacy extensions. So implementing this change may be difficult.
to
In practice, each operating system has a different way of indicating if an IPv6 address comes from a privacy extension or not. Usually the operating system also returns the state of each available address: "public" - the ones that does not change, and which Tor should use "temporary" - the ones that come from privacy extensions Tor should always ask for and use a "public" IPv6 addresses to build relay descriptor.
What's the downside of using "temporary" IPv6 addresses from privacy extensions?
I mean, isn't better privacy a good thing?
<SNIP>
Not really. These privacy extensions IPv6 addresses might be good for outbound bind exit addresses (for Exit relays), and maybe (not sure) for regular clients that could connect to their entry guards or bridges using a temporary IPv6 address.
We only refer in this proposal to Tor in _relay mode_. When in relay mode, it is desirable to bind to a static IPv6 address that does not change, so bandwidth authorities can measure its bandwidth and directory authorities and maintain its history, uptime statistics and flags as well as not upload descriptors too often that will make them unusuable for clients that have an older consensus which is still valid, and so on.
Usually it is not desirable for a 'server' of any kind (Tor relay included of course) to have an expiring / temporary / dynamic IP address. It is the other way around actually.
So, we don't plan to throw poison on privacy extensions IPv6 addresses, might actually use them for the purposes explained at the beginning of this email, but in this particular case of Proposal 312 when we are discussing automatic address discovery for *relays* they are bad for us - we wouldn't want to code Tor to discovery and gladly use a temporary IPv6 address that was designed to *not* be used in server mode.