RE: What is the purpose of the discussion? Answer

[Burke Chris-CCB007 <Chris.Burke@motorola.com>]

Paul Reynolds wrote (in part):

>The percieved problem is that the IETF needs to tackle, that isn't already
>within  scope for an existing WGs, is to have protocol development more focused
>upon wireless requirements...

draft-reynolds-mobile-isp-requirements-00 describes a system called a "wireless internet framework" that presumably offers (reproduces?) the functionality of a digital cellular network, using internet technologies.

Thought exercise.

One could write a similar draft, describing a system called a "laundry internet framework" that reproduces the functionality of a commercial laundry, using internet technologies. You see, the IETF has historically not taken the requirements of these huge laundry systems, which serve hundreds of millions of people worldwide, into account when designing protocols. A few simple examples: protocols are needed that account for the proper pH balance of the washing solution based on load type; a billing system is needed since laundry isn't done for free; QoS must be assured throughout the system since goods may be damaged by incorrect cycle or bin transfer timing.

When considering the hypothetical laundry internet framework, or LIF, draft, what would the "next step" be? One approach would be to separate the requirements into four bins: (1) application-specific or low-level infrastructure-specific requirements having nothing to do with the IETF; (2) requirements that can be satisfied by existing internet practice; (3) requirements having to do with proposed internet features that apply across a range of systems (not just laundry systems); (4) requirements having to do with new internet features specific to the laundry application.

As a natural consequence of their industry-specific perspective, the hypothetical authors of the LIF draft would tend to place many "requirements" in bin (4), that the internet community might deal with more effectively via one of the other bins. (We all know the importance of properly sorting the laundry...). One of the keys to the success of the Laundry Internet Framework, then, might be to overcome this tendency and appropriately sort the requirements.

End of thought exercise.

The text of draft-reynolds-mobile-isp-requirements-00 presents a set "Wireless Internet Framework" system requirements ("needs to..."). I believe the draft succeeds at describing key attributes of a system with surface behavior like that of a digital cellular network, but with IP internals. I also believe that many of the requirements listed in the draft are not at all wireless specific. The requirements that apply to any future commercial internet include:

- Requirements related to billing for services
- Requirements related to offering services across multiple physical bearers
- Requirements related to offering services across multiple administrative domains
- Requirements related to creating a "walled garden" of services
- Requirements related to QoS and, more generally, collaborative system resource allocation
- Requirements related to security, privacy, fraud prevention, and lawful intercept
- Requirements related to system performance monitoring and reliability

As a person who's been in the wireless data industry for nearly 20 years, I see genuine system-specific requirements in many of these categories, but I don't see any of them requiring a wireless-specific change in the overall architectural philosophy of the internet. My guess is that many of these requirements can be satisfied 90% by solving the related problem for the internet at large; this is the problem set I think the IETF should address.

The set of truly wireless-specific technical issues is rather small. Here's the basis set:

1) Micro-mobility - Some wireless nodes move around; some change their network point of attachment multiple times within a session; some use time-variant / intermittently viable link layers for which packet loss does not imply congestion; some use asymmetric link layers; some use relatively slow (2 or more orders of magnitude below wired internet speeds) link layers (PILC wg, Mobile IP wg, IPNG wg stuff)

2) Economic assumptions - Some wireless nodes make use of scarce, expensive communications bandwidth (e.g. $0.10 charged to the user for every kilobyte of data transfer; $0.30 charged to the user for every minute of a session); some have very limited compute capability, storage capability, display capability, data entry capability, etc (factors encouraging specific types of protocol and application optimizations including application layer gateways like WAP)

3) Regulatory environment - Some wireless nodes operate in a different regulatory environment than the U.S. conception of the wired internet. As wireless devices become less telephone-like and more internet-host-like, the response from global regulators ranges from reactionary to insightful. Regardless of the specific regulatory reaction, typically vendors and service providers of wireless devices operate under regulatory constraints that imply (sometimes ludicrous) technical "requirements" for wireless nodes.

We might go a long way toward addressing the genuine concerns of the wireless industry, if the IETF and MWIF were to focus on defining a "memo of understanding" on which parts of the problem are appropriate IETF issues and which belong in MWIF. The objective of the MoU would not be to carve up turf, but rather, to set expectations, avoid misunderstandings, and speed progress toward solutions.

Chris Burke