Re: Spin bit as a negotiated option

[Kazuho Oku <[e-mail]>]

2018年10月5日(金) 6:55 <[e-mail]>:
>
> On 10/04/18 23:40, Kazuho Oku wrote:
> > 2018年10月4日(木) 19:46 <[e-mail]>:
> >>
> >> Thanks Kazuho. Using version numbers directly was an obvious choice, but so far
> >> I was discouraged by strong language in the spec, saying 0x00000001 was the
> >> unavoidable point of convergence after all 0xFF0000XX experiments. So in essence
> >> you're proposing to weaken that a bit, right ?
> >
> > I am not sure if we have reached consensus that QUICv1 should use
> > 0x00000001 as the one and only number for the version number field.
> >
> > My understanding is that it is simply written as such because I-D is a
> > draft of a RFC. A RFC needs to clarify the value of version number
> > field it uses, and I-D needs to have a paragraph that amends on how
> > the draft-versions uses the field.
>
> OK, thanks for the clarification.
>
> > I do not think we can segment the version number space, because IMO it
> > essentially means making spin bits an invariant. My understanding is
> > that spin bits is something specific to v1.
> >
> > QUICv1 might use different values on the version number field to
> > represent if spin bits is (or other signals are) available, but that
> > does not mean that v2 will expose the same signals or they would use
> > the version number field for indicating that.
> >
> > If it is the case that the use of 0x1xx for V1 might give people the
> > impression that the lower bits are used to indicate the availability
> > of unencrypted signals across multiple versions, I think starting from
> > 0x00000001 (i.e. the one you describe as "direct enumeration") would
> > be preferable.
>
> Crystal clear, and agreed:)
>
> >> Also, are we sure that this explosion won't weigh on the new VN scheme, inducing
> >> slower convergence due to longer lists on either side ?
> >
> > I do not think that is an issue for v1, because what currently at
> > stake is if we should assign one or two extra version numbers (i.e.
> > one for spin bit, and possibly another for spin+vec as a separate
> > draft).
> >
> > If folks agree to expose other signals that requires a "configuration"
> > to be exposed as well, then they can define a new version number that
> > introduces a new field to the long header format for exposing the
> > "configuration."
>
> OK. So, wrapping up, and abstracting away for now the bit layout of the version
> number by using symbolic names, we'd start (right after Bangkok) with something
> like:
>
>         - "v1"     = v1 without extensions
>         - "v1s"    = v1 with spin bit only
>         - "v1svv"  = v1 with spin bit and VEC
>         - "v1sqr"  = v1 with spin bit and loss bits Q and R
>         - etc (?)

Yes.

We will have "v1" and "v1s", assuming that we would agree to ship the
one-bit "spin bit" approach as part of QUICv1. People will also have
the freedom for using more bits for the flavors (e.g. "v1svvv",
"v1sqr").

>
> One question though: what part of the version negotiation exchange is in clear ?
> IOW, would the selected version always be exposed to the path, or should
> spin-consumers rely on heuristics to guess the variant in use ?

My understanding is (assuming #1755 gets merged):
* the version number field of a long header would always be "v1" for
the first Initial sent from the client
* the value would be the negotiated version for exchanges that follows that
* there is no version number field in the short header

Therefore, for each path identified the 5-tuple, an observer would
remember the version number of the last long header packet that it has
seen (or it could remember the version number of any Handshake
packet), and use that to determine what is being exposed on the wire.
No heuristics is required.

OTOH, because version number field does not exist in the short header,
an observer cannot determine if any signal is exposed for a connection
after it migrates to a 5-tuple, or for a connection that has been
alive before starting the observation. Use of heuristics would be the
only way for such cases.

>
>
>
>
> _________________________________________________________________________________________________________________________
>
> Ce message et ses pieces jointes peuvent contenir des informations confidentielles ou privilegiees et ne doivent donc
> pas etre diffuses, exploites ou copies sans autorisation. Si vous avez recu ce message par erreur, veuillez le signaler
> a l'expediteur et le detruire ainsi que les pieces jointes. Les messages electroniques etant susceptibles d'alteration,
> Orange decline toute responsabilite si ce message a ete altere, deforme ou falsifie. Merci.
>
> This message and its attachments may contain confidential or privileged information that may be protected by law;
> they should not be distributed, used or copied without authorisation.
> If you have received this email in error, please notify the sender and delete this message and its attachments.
> As emails may be altered, Orange is not liable for messages that have been modified, changed or falsified.
> Thank you.
>


-- 
Kazuho Oku