[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: queries for unbounded recurring components



John Stracke wrote:
> 
> Doug Royer wrote:
> 
> > Bruce_Kahn@iris.com wrote:
> >
> > > If the data is encrypted, then unless the CS has the key to crack open the
> > > data and return the expected chunks we have to deal with another serious
> > > architecture issue.  The same issues applies to the extracting a single
> > > instance from an encrypted chunk and then reencrypting it for the CU; just
> > > what do we expect the CS to realistically do?
> >
> > If the CS does not have the key, it could not have returned ANY of the data.
> > And the CS would not know how to return any date ranges. So that will not
> > work.
> 
> But, for *signed* data, Bruce's
> objection still stands. 

Had he made it :-)

> It ought to be possible for someone to submit an S/MIME
> message with a text/calendar body, and have the CS store the whole message.  The
> CS may or may not check the signature itself, but it can then send the signed
> message down in response to queries, so that clients can check the signature if
> they want--that way they can do end-to-end authentication, and not have to trust
> the CS not to make things up.

True - assumes the CS does not change anything in the object- and it can
depending on the application.  So as long as it can strip out the
signature (your #1 below). 

If an ATTENDEE sends a RSVP yes, and the CUA or a bot updates the status
information - then the signature of the original object would have to
be tossed.

> If we do this, it raises a couple of further questions:
> 
>   1. Should the CS, instead of having to store the whole message, be allowed to
>      strip out the text/calendar from the S/MIME, discarding the signature?
>   2. If so, should the CS be able to tell the sending CUA not to bother sending
>      S/MIME?
>   3. Should the receiving CUA be able to specify that it doesn't want the
>      signature?
> 
> ...
> On #2, again, I think the answer is yes,
> since it saves the CUA from wasting bandwidth and CPU time, and also tell the
> user that the server can't store signed data.

If the message arrived into an MUA (iMIP) and then dropped into a CUA - 
what good would it do, you have a signature anyway?

A CUA is responsible for formatting the object before sending it down
to a CS. If it does not like the bandwidth or speed of the connection
then that is a CUA issue - not a CAP issue as the CUA does not have
to send one down.

There is also no requirement that a CS support encryption or
digital signatures. So instead on saying you can't do it, maybe
there needs to be a capability saying a CS does sore signatures.
And if the capability is not present, then there is no way to tell.

> On #3...probably yes, but I'm not
> as sure about it; the savings are less (a CUA that doesn't verify signatures
> doesn't spend much CPU time on ignoring them), and the complexity may not be
> worth it.

Just as with email. Clients just ignore them.
begin:vcard 
n:Royer;Doug
tel;pager:650-274-8960 or pager@Royer.com
tel;cell:650-274-8960
tel;home:650-274-8960
tel;work:805-957-1790 x541
x-mozilla-html:FALSE
url:http://Royer.com/People/Doug
version:2.1
email;internet:doug@home.royer.com
adr;quoted-printable:;;801 Woodside Rd. #14=0D=0ASuite 244;Redwood City;CA;94061;USA
x-mozilla-cpt:;0
fn:Doug Royer
end:vcard

Attachment: smime.p7s
Description: S/MIME Cryptographic Signature