Re: MPLS Inter-area TE requirement draft

Jean Philippe,

> At 11:55 AM 1/2/2004 -0800, Yakov Rekhter wrote:
> >Jean Philippe,
> >
> > > Jim,
> > >
> > > At 07:05 PM 12/31/2003 -0800, Jim Boyle wrote:
> > >
> > > >JP, so you state that both optimality and scalability are
> > > >requirements, yet you acknowledge that there will be trade-offs.  I
> > > >believe it is important to prioritize the requirements, so as to best
> > > >guide the discussion on the solution.
> > >
> > > A few comments:
> > > - you seem to make the statement that optimal always means non scalable,
> > > something I disagree with. Of course, more optimal very likely means more
> > > expensive to compute, hence the trade-off I was referring to.
> >
> >Whether optimality means scalability depends on the problem complexity.
> >E.g., for problems with log or (low) polynomial complexity optimality
> >need *not* mean non scalable; on the other hand, for NP-complete problems
> >optimality certainly *does* mean non-scalable.
> 
> fully agree. Here, the potential solution one has in mind (PCS) relies on 
> CSPF, although more sophisticated algorithms could be used of course.
> By optimal we mean as optimal as in the case of a single area (shortest 
> constrained path).

In this case what you call "optimal" is really a *local* optimal.
So, the question to ask is whether requirement is to produce a
locally optimal solution only, or whether the goal is produce a
globally optimal solution.

Yakov.