Re: [Roll] [roll] #5: DODAG

[Philip Levis <pal at cs.stanford.edu>]

On Nov 9, 2009, at 10:30 AM, Julien Abeille (jabeille) wrote:

> Hi Phil,
>
> > -----Original Message-----
> > From: Philip Levis [mailto:pal at cs.stanford.edu]
> > Sent: lundi 9 novembre 2009 18:12
> > To: Julien Abeille (jabeille)
> > Cc: Tim Winter; ROLL WG
> > Subject: Re: [Roll] [roll] #5: DODAG
> >
> >
> > On Nov 9, 2009, at 2:11 AM, Julien Abeille (jabeille) wrote:
> >
> > > Hi Tim,
> > >
> > > My question is: are there DAGs that are not DODAGs in RPL?
> > My feeling
> > > is no. Hence I would suppress the DODAG instead of turning
> > all "DAG"
> > > into "DODAG".
> > > Best,
> > > Julien
> >
> > I disagree -- Directed Acyclic Graph is a very well defined
> > term in graph theory. We shouldn't redefine it. A DODAG is a
> > particular kind of DAG, where there is one and only one node
> > which has only incoming edges. We could maybe come up with a
> > better term than DODAG, but DAG definitely isn't it. It will
> > confuse readers tremendously.
> Sorry, did not want to hurt graph theory specialists among IETF draft
> readers.
>
> Seriously, what is confusing is to have DAGs and DODAGs everywhere in
> the draft.
> An instance is not as far as I understand a DAG. An instance is  
> multiple
> DODAGs. Hence either we use only DAG everywhere, or DODAG  
> everywhere, no
> mention of DAG. Thoughts?

I totally agree - the mixed use of DODAG and DAG is confusing. I think  
we should use DODAG.

Phil