Re: [tor-dev] Marking Proposal 299 (Preferring IPv4 or IPv6 based on IP Version Failure Count) as Accepted

Hi Neel,

On 17 Feb 2019, at 08:10, neel@xxxxxxxxx wrote:

My proposal "Preferring IPv4 or IPv6 based on IP Version Failure Count" (a.k.a. Prop299) is here: https://gitweb.torproject.org/torspec.git/tree/proposals/001-process.txt

That's a link to the proposals process document.

The proposal is here:

https://gitweb.torproject.org/torspec.git/tree/proposals/299-ip-failure-count.txt

I haven't gotten any comments for requests when I asked for them, so I am assuming this proposal is okay.

Here are the tor-dev threads responding to your proposal:

The January thread starts here:

https://lists.torproject.org/pipermail/tor-dev/2019-January/013648.html

I also did an in-depth review in February:

https://lists.torproject.org/pipermail/tor-dev/2019-February/013673.html

If it is, could someone please mark this proposal as Accepted? If not, what does this proposal require?

Here's what Accepted proposals require:

   Accepted: The proposal is complete, and we intend to implement it.
      After this point, substantive changes to the proposal should be
      avoided, and regarded as a sign of the process having failed
      somewhere.

https://gitweb.torproject.org/torspec.git/tree/proposals/001-process.txt#n152

But in my detailed review, I said:

Here's one thing we must fix before we start implementing this proposal:

We don't store connection statistics on Tor clients right now.  This proposal would make us store these statistics.

Then I suggested some different ways to avoid collecting connection statistics on clients.

Here's what you could do:

Change the proposal so that it doesn't collect user connection statistics.

Write some code, and do some testing to answer the other questions:

I think these questions are minor tweaks, so they can be changed after the proposal is accepted:

1. What is the starting SFPV?

These questions may require a significant re-design, so we need to answer then before the proposal is accepted:

2. When switching between IPv4-only and IPv6-only networks, the circuit failure rate could start as high as 87.5% (7/8) … What does the pathbias code do when this many failures happen?

3. What happens on a network which drops IPv4 or IPv6 packets?

4. Do we want to count successful connections?

For example, to fix 2 and 3, we might need to make sure that there is at least one IPv4 and one IPv6 connection in every N pending connections. 

Other people may also have specific questions that need to be answered before the proposal is accepted. (I looked in the initial thread, and I couldn't find any.)

T