Re: [quicwg/base-drafts] disable_active_migration with SPA (#3765)

[Igor Lubashev <[e-mail]>]

To summarize, it seems that people are generally not philosophically opposed to fixing this issue, but the solution needs to be simple and be compatible with the current behavior.

@MikeBishop has a PR that has one point unresolved -- what to do about SPA that is of a different protocol family.

I would like to propose a simple but workable solution.  _When "disable [active] migration" bit is set on the SPA, the client has to use the handshake source address if it is of the same address family as the SPA address it is migrating to.  If SPA is of a different address family, the client will have to change the source address but then stick with it for the duration of the connection._


If the server's need to prohibit active migration with SPA is due to the client's need to keep the same source address, the server will only return an SPA of the same address family and set "disable [active] migration" bit.  If the server is only concerned with the client staying on the same network, it can return a different-family SPA (and set "disable [active] migration" bit), since the chances are excellent the client will stay with the same network when it migrates very soon after the handshake completes.  

(In case the client tries to migrate to another address family and ends up on an unacceptable network, the server can probe the original path. If the original path is valid, the server would ignore packets coming on the new path, and the client would fail back to the original path.  If the original path is no longer valid, the server would close the connection on the new path.  The result would be a hiccup for the client, but this would happen rarely.)

-- 
You are receiving this because you are subscribed to this thread.
Reply to this email directly or view it on GitHub:
https://github.com/quicwg/base-drafts/issues/3765#issuecomment-665386732