Re: [Roll] some comments on draft-thubert-dao-projection-00.txt

["Pascal Thubert (pthubert)" <[e-mail]>]

Hello Michael

> 
>     > Per RFC 6550, a path constructed by storing-mode DAO messages is
>     > directional, packets flow parent to child, and we do not want to
>     > introduce confusion by changing any of that.  Thus the segment that is
>     > constructed from the segment egress to the ingress.
> 
> I don't think I understand what this is an alternative to.
> I.e. what would the way to do it wrong be?  Why are you suddenly talking
> about storing-mode DAO messages.

I did that because the signaling of the projected route being installed is hop by hop using DAO in a fashion that extend storing mode.
The DAO flow from egress to ingress of that projected route while the data flow the other way. 

> 
>     > It results that the segment is directional but not that it is
>     > necessarily flowing down the DADOG, this depends on how much
>     > information is available to the root or whatever PCE is used as a
>     > helper.
> 
> This seems to imply that traffic could flow across the mesh... i.e. hop
> From branch to branch.   How would the DODAG root know that this was
> possible
> using pure RPL?  I think that it couldn't without 6top or something.

Yes.  DetNet may provide means for that.
The draft does not say where the root could get more information but it does not preclude the projecting routes across the mesh.

Cheers, 

Pascal