The IP over ATM Mailing List Archive by date

[Curtis Villamizar <curtis@ans.net>]

In message <199505120742.QAA17051@necom830.cc.titech.ac.jp>, Masataka Ohta writ
es:
> > There is no reason why a LIS should be made large.  The fact that LIS
> > are kept small and interconnected by routers to form an internetwork
> > is a feature of IP, and a featue of Classical IP over ATM, not a bug.
> 
> If you are content with packet-wise relaying by routers, we don't
> need ATM subnet logical. So, make LIS physically contegeous entity
> separated by IP routers and call LIS just IP subnet.
> 
> But, the current framework document draft says:
> 
> :IP over ATM should therefore be characterized by:
> 
> :  o Degree  to which a  connection oriented  lower level is  available
> :    and utilized.

This is described as a characteristic.  Nothing more.

Nowhere in this section does it say whether using a more connection
oriented approach is better or worse.  If you look in section 5, both
sides of the argument are acknowledged with an attempt not to take
sides.

> > > > LIS will be fine.  But, LIS will have some issues (e.g. "complete" 
> > > > autoconfiguration). 
> > > 
> > > You had better call your geographically crippled version of LIS
> > > PLIS (Physical Logical IP Subnet) or, more simply, just subnet.
> 
> > Yes.  LIS equals subnet.  The Classical model does not try to scale a
> > LIS to global scale, so there is no sense criticizing it for not
> > acheiving a non-goal.  The Classical model works very well and there
> > is no reason to refer to it as crippled.
> 
> No, the classical model does not try to scale a number of hosts in
> a LIS to global scale. So what?

OK.

> The current framework document draft also says:
> 
> :The following characteristics of the  IP over ATM internetwork may  be
> :independent of geographic dispersion (LAN, MAN, or WAN).
> 
> :  o The size of ATM IP subnets (LIS) in the ATM Internetwork.

The point of this introductory section is that theses are independent
characteristics.  A WATM does not have to mean big LIS and an LATM
does not have to mean small LIS.  No statement is made at this point
in the document as to feasibility of protocol scaling.

Again, with respect to scaling, attempt is made to represent multiple
opinions expressed by this work group.  The introductory section you
point to does not cover scaling issues, it simply enumerates
characteristics of ATM independent of the IP over ATM models.

Questions of scaling are brought up in Section 6.

> > We know that you are working on an alternative and there is a separate
> > WG to discuss that alternative.
> 
> There are no WGs yet. Moreover, we are discussing on the objective of
> IP over ATM protocols, in general, not related to a specific proposal.

Fine.  There is still a separate mailing list for colip.

Being more "connection oriented" with no intermediate hops is not an
explicit goal of the Classical IP over ATM proposal nor is it a
specific goal of at least some IP over ATM customers.  Scalability is
a goal and support for integrated services including both elastic and
real-time service is a goal.  I feel that the Classical IP over ATM
proposal with RSVP support can meet these goals.

I don't feel that it is appropriate for me to state my opinion as fact
or the consensus opinion in the framework document, so all proposals
are represented and the issues enumerated, but none are singled out as
best or worst.

> > ps- Please let this thread die.
> 
> If you insists that packet-wise relaying is enough, you should let
> the IPATM WG die, instead.

Why.  We expect to be buying ATM service and equipment and want IP
over ATM to work (very well if possible).  I don't see "being
connected oriented" as an explicit goal and see no reason to abandon
IP over ATM for that reason.

> 							Masataka Ohta

Curtis