--- Dan Merillat <[EMAIL PROTECTED]> wrote: 
> On Wed, 24 Sep 2003, Some Guy wrote:
> 
> > > DPK@<keyring>/url
> 
> > Instead of coming up with a new key type and
> having
> > even longer URLs, would it be possible to just
> stick
> 
> > with SSK format but just sign the keyring once
> with
> > the private key and then throw it away.
> 
> good god NO. There's no way to know it's been
> "thrown away"

The SSK URL format could have been saved, if you used
this throw away technique, but yes I understand your
objection.  You don't want to trust anyone more than
the rest of the signers even for a second.

Anyway, it seems if you come up with some new
authentification, it will provide a superset of the
functionality of SSK, so eventually SSK might be
redundant.

> > This data should be routed to the same place as
> the
> > rest of the data.  I see no reason why that's not
> > possible.
> 
> Clumping is bad.  We don't want to introduce a way
> to force
> an unlimited amount of "related" data to one node.

This data is very related!!!  In fact, neither the
data nor the certificates are of any value by
themselves.  The node with the data needs the
certificates before it can even upload the data one
hop.  I'm sorry, but I see no reason why they should
be stored somewhere else.  The node caching the data
needs them to validate an insert and to upload for a
request.  Why should it throw the thing somewhere out
there to retrieve later?


> If done this way, it's an even easier attack then
> the 10-15 
> bit flooding attack I was discussing earlier. 
> Generate random
> keyrings until you find one in the appropriate
> keyspace, insert it,
> and sign lots of random data with it.  (Under 32k,
> same as a SSK)
Ahh, insteresting point about the SSK DNS; I didn't
think of that.  There's nothing We can do about that
though is there?  

The DNS issue is important.  You have a choice:
A) All certificates have to be handed in together, so
they can be checked on the way to the target.
B) A mutually annoymous group can insert thier
certificates seperately.

A will solve most our problems, and resists the
possible floods from rouge members.

B is cooler, but it may have to have something to
prevent a rouge member of the ring, bombing nodes, by
making junk certificates.  Something like a limited
number of certified slots,  could be cached so the
rouge cann't resend indefinately. 

Or wait let's combine our ideas!!!
Store the individual signatures in different places,
but to insert actual data, one first must collect
those signatures and the policy to send with the data
so you can validate it on it's way in.  This would
allow the individual signers only to DNS thier own
signature.  This seems like the best you could do. 
It'd still be possible to do a seperate DNS on the
hash space using a seperate SSK attack though.

> > Not if you use my suggestion.  You could possibly
> make
> > it to where if "9 of 10 want to change the policy,
> > they can."  This could lead to a string of
> > certificates required on insertion, but otherwise
> the
> > URL wouldn't change.  Implementing this may be
> pain in
> > the butt.
> 
> That was my suggestion, actually.  Please re-read
> the proposal
> again.
I was talking about my suggestion with the throw away
key.  I'm just tring to come up with a shorter key
format so you can have 100 signers without a 5000
character URL.

Here's better idea to keep the keys short.
URL=
HASH(<key1><key2><key3>..<keyn>policy_blablabla)/page.html
This should keep the URLs from exploding.  Sound ok?


Cool discussion!!!

__________________________________________________________________

Gesendet von Yahoo! Mail - http://mail.yahoo.de
Logos und Klingelt�ne f�rs Handy bei http://sms.yahoo.de
_______________________________________________
Devl mailing list
[EMAIL PROTECTED]
http://dodo.freenetproject.org/cgi-bin/mailman/listinfo/devl

Reply via email to