That was part of the plan for the 0.3 spec.

Remote sends delta.
Host replies with error saying it needs the signature
Remote sends signature+delta.

On Sun, Nov 8, 2009 at 13:19, Ben Kalman <[email protected]> wrote:
> Yes, that is the plan.  It would even (perhaps) be nice to modify the
> protocol so that signer info can optionally be sent with an XMPP submit, to
> reduce the number of necessary roundtrips.
>
> On Sun, Nov 8, 2009 at 1:14 PM, Tad Glines <[email protected]> wrote:
>>
>> OK, I should have read your previous post more carefully.
>>
>> You're saying that the host will simply reject a delta if it doesn't
>> have the signer info, thus forcing the remote to send it and then
>> re-transmit the delta. That will work. It also reduces the amount of
>> state the host has to keep track of. That's something I hadn't
>> considered.
>>
>> On Sat, Nov 7, 2009 at 5:58 PM, Ben Kalman <[email protected]> wrote:
>> > As I said before, the host of a wavelet should not have to deal with
>> > certificate distribution itself, it should just accept and distribute
>> > updates to wavelets.  This is why the host will not do a
>> > getDeltaSignerInfo.
>> >
>> >
>> > On Sun, Nov 8, 2009 at 12:54 PM, Tad Glines <[email protected]>
>> > wrote:
>> >>
>> >> Sending the same certificate chain with each delta is a massive waste
>> >> of bandwidth.
>> >> There is no reason to send the same exact chain each time. If the only
>> >> way to get a cert chain was via getSignerInfo then server's would only
>> >> need to ask once for each new signer info id encountered.
>> >> Yes there is a little extra delay with that first delta containing a
>> >> new signer info id, but it'll save on lots of bandwidth later.
>> >>
>> >>
>> >> On Sat, Nov 7, 2009 at 5:38 PM, Ben Kalman <[email protected]> wrote:
>> >> > Whenever a server send a signed delta (whether that be due to
>> >> > broadcasting
>> >> > an update as the host, or submitting a delta as the remote) the
>> >> > receiving
>> >> > server must be able to verify the delta, so must have the
>> >> > corresponding
>> >> > certificate (signer info) for the signer id encoded into the signed
>> >> > delta.
>> >> >
>> >> > Whether the server gets this through being posted the signer info
>> >> > (postSignerInfo) or whether it has to request it (getDeltaSignerInfo)
>> >> > depends on where the wavelet is hosted -- it is up to the remote
>> >> > server
>> >> > to
>> >> > ensure that all the certificates in place, i.e. the host server
>> >> > should
>> >> > never
>> >> > have to worry about distribution of certificates.  So the remote must
>> >> > getDeltaSignerInfo if it receives an update with a missing
>> >> > certificate,
>> >> > and
>> >> > postSignerInfo on every submit.
>> >> >
>> >> > So, two separate issues with this: previously (up until a9597f31ff)
>> >> > there
>> >> > was a rather bad shortcut taken to avoid the complications of
>> >> > callbacks/queuing (tight time schedule :), where certificates were
>> >> > posted to
>> >> > remove servers from the host on updates.  This has been fixed.
>> >> >
>> >> > The other issue is (as Tad noticed) a certificate is posted with
>> >> > every
>> >> > submit.  This is necessary based on an assumption that the host
>> >> > doesn't
>> >> > have
>> >> > our certificate, since we have no way of knowing this.  It's also
>> >> > somewhat
>> >> > realistic given that FedOne servers don't persist certificates.
>> >> > However,
>> >> > once we formalise an error spec (soon) and FedOne/Google supports
>> >> > propagation of error messages/codes (soon), the plan is for the host
>> >> > to
>> >> > reject signed deltas with are missing signer info with a recognisable
>> >> > error
>> >> > code, so that the remote can then send the signer info only when
>> >> > required.
>> >> >
>> >> > Hope that made sense :-)
>> >> > -- Ben
>> >> >
>> >> >
>> >> > On Sun, Nov 8, 2009 at 7:20 AM, Daniel Danopia <[email protected]>
>> >> > wrote:
>> >> >>
>> >> >> This is funny because Sails hasn't gotten a single certificate in
>> >> >> over
>> >> >> a week from FedOnes, since FedOne requests them now xD
>> >> >>
>> >> >> On Nov 6, 8:28 pm, Tad Glines <[email protected]> wrote:
>> >> >> > It looks like FedOne will post signer info to the fed host every
>> >> >> > time
>> >> >> > it sends a delta.
>> >> >> > This seems very inefficient. It seems to me that FedOne should
>> >> >> > only
>> >> >> > send signer info as a result of a direct request.
>> >> >> >
>> >> >> > I thought I might have seen an issue or code review request
>> >> >> > related
>> >> >> > to
>> >> >> > this, but I couldn't find it.
>> >> >> >
>> >> >> > -Tad
>> >> >>
>> >> >
>> >> >
>> >> > >
>> >> >
>> >>
>> >>
>> >
>> >
>> > >
>> >
>>
>>
>
>
> >
>



-- 
Anthony Baxter, [email protected]

--~--~---------~--~----~------------~-------~--~----~
You received this message because you are subscribed to the Google Groups "Wave 
Protocol" group.
To post to this group, send email to [email protected]
To unsubscribe from this group, send email to 
[email protected]
For more options, visit this group at 
http://groups.google.com/group/wave-protocol?hl=en
-~----------~----~----~----~------~----~------~--~---

Reply via email to