Re: Alternate Version + Salt in Alt-Svc

["Martin Thomson" <[e-mail]>]

On Tue, Nov 5, 2019, at 07:34, Ryan Hamilton wrote:
> I can think of solutions/mitigations for this issue (like restricting 
> the number of bits which can be flipped by the server, or expiring the 
> alternative version in the client on first use, etc) but I'd be curious 
> if this seems like an issue to other folks first.

Yes, this is an issue.  Thanks for identifying it.  We hadn't gotten to Alt-Svc for this yet, we're having enough trouble with NEW_TOKEN :)

The general rule we follow for these things is to require confidentiality for the initial signaling and to only use the value once.  That works for session tickets, it works for NEW_TOKEN, and I see no reason not to have it apply here also.  That means falling back to version==1 for connection attempts after the first.

That diminishes the value of the addition a little, but I don't think that it is enough to be concerned.  The alternative service still exists and can be used, and successful connection attempts should provision the client with more tokens for future connections.