Re: [Demexp-dev] Some definitions of delegation

[David MENTRE]

Hello Félix,

Many thanks for your review.

Félix <address@hidden> writes:

> I read your paper. Concerning note 12, I would not say that the whole
> chain is removed. Only the "first" delegation should be ignored. In other
> words, if I refer to the syntax of definition 6, only "p delegates on
> d'(1) for tag t' "
> because the rest of the chain should be left unchanged (other participants
> are "using" this chain!)

You're right. I've thought about it and arrived to the same conclusion.


> I hope I'm clear enough (cannot use the nice arrows you used in the
> document:-)
> Now, in terms of algorithms, one cannot simply delete the "week" delegation
> because it is still valid for questions having the tag t' and not having
> the tag t.

Yes, a good example of issue to solve is as followed:

i---t,t'->-
           \   -t--> d_A .....> q_1 : t
            \ /         .
             d           ......
            / \                .
           /   -t'-> d_B .......> q_2 : t,t'
i'--t,t'->-

[ i and i' delegates to d for tags t and t'. d delegates to d_A for t,
  d_B for t'. d_A votes on question q_1 having tag t. d_B votes on
  question q_2 having tags t & t'. ]

In that case, we want to solve the conflict for q_2 but still keep
delegation non-conflicting delegation for q_1.

> Therefore maybe the expression "and p----t'--->d' is removed" that you used
> in note 12 should be modified into "and p----t'--->d' is not taken into
> account
> for this questoin" or something equivelant.

Yes, I'm trying to refine me definitions. I'll submit a new
proposal. I'm thinking about a kind of Conflict Resolution Table that
indicates the delegate to take into account in case of conflict
(e.g. d_A if for i  t  > t'
      d_B if for i' t' > t ).



Yours,
david
-- 
pub  1024D/A3AD7A2A 2004-10-03 David MENTRE <address@hidden>
 5996 CC46 4612 9CA4 3562  D7AC 6C67 9E96 A3AD 7A2A