What is the purpose of a standard?
“The focus of standards development should be the benefit of humanity.” Andrew Myles (quoting from the IEEE’s mission statement).
Andrew Myles has been involved in standards development for decades, starting in the late 1980s. He has been a contributor to ANSI’s FDDI development (ANSI X3T9.5), to the IETF’s Mobile-IP development group, a very long time contributor to IEEE’s 802.11 working group, to 3GPP’s RAN1 Technical Specification Group subcommittee (which focuses on the physical layer for cellular radio interfaces) as well as ETSI’s Broadband Radio Access Network (“BRAN”) committee to help harmonize standards for 5GHz and 6GHz and promote fair coexistence between Wi-Fi, cellular and other radio based telecommunications standards. He has been Chair of the Board of the Wi-Fi Alliance and a member of the Institute of Electrical and Electronics Engineers (“IEEE”) Standards Association (“IEEE-SA”) Board of Governors. He is currently a member of the IEEE-SA Standards Board.
Andrew received his PhD from Macquarie University in Australia – his thesis was on “Protocols for mobile networks at the MAC and Network Layers.” A few years after receiving his PhD, Andrew joined Radiata Communications, one of the first companies dedicated to making 802.11a Wi-Fi chips (Radiata was founded by two Macquarie professors). Cisco Systems acquired Radiata in February 2001. I met Andrew shortly after I arrived at Cisco in 2005. Andrew recently retired from Cisco, but continues to be involved in standards. You can read more about Andrew’s many accomplishments on his LinkedIn page: https://www.linkedin.com/in/andrew-myles/.
I recently had the privilege of reconnecting with Andrew. What follows is a summary of our wide-ranging discussion on standards development and governance, what Andrew has seen change over his many years of being involved in such development and what he perceives would be helpful for future standards development.[1]
Let’s start with the basic question of what is “a” standard. I ask this question because sometimes, for example, we say the 802.11 standard or the Wi-Fi standard. Other times, we say the 802.11n or 802.11ac standard. What is the difference between these things?
I am going to answer this question in the context of the “802.11 standard.” The IEEE 802.11 standard is a standard that specifies the PHY (physical) and MAC (media access control) layers for a wireless local area network (WLAN), with a focus on the interfaces required for interoperability. The specifications in the 802.11 standard allow implementors to build devices that work together as part of a high-performance WLAN. The IEEE 802.11 standard is a single document that has had numerous updates and amendments since its original development way back in the 1990s.[2]
For 802.11, there is a main branch of amendments which include the amendments that are named and generally known by the public, such as 802.11n and 802.11ac. But there are other branches of the development effort that result in amendments to the standard that focus on smaller, less integral aspects of the standard or aspects that are not continually being improved. Sometimes these side branches become quite important to the overall success of the standard. For example, 802.11i, which is known as Wi-Fi Protected Access (WPA and WPA2), patched up a large security hole (aka the “WEP debacle”) in the early 2000’s and was one of those vital side branches.[3]
Every so often, we take the base standard and all the amendments that have come out since that base standard version was released and roll it up into an updated base standard. But, there is only one standard – IEEE 802.11 - with different amendments and updates that have been released over time.
What makes a standard successful?
A standard is a piece of paper which is not in and of itself very useful or interesting. It is important to remember that most standards are not very successful. To be successful, it takes good features, consensus and a bit of luck. The trick is to develop a useful standard and to get a lot of implementers to actually implement it. The industry and the implementers within that industry ultimately determine whether a standard is successful or not.
Many of my readers are lawyers and do not have much experience with the nitty-gritty of standards development. Can you give us an overview of how a standard is developed.
Developing a standard is a process. Very rarely are you starting from scratch. For example, when I started working on the 802.11 standard, there already was an existing standard with several iterations.[3] I’ve been involved with development of the 802.11 standard for a long time, so I’ll use that as the baseline for this overview.
There are a few different things going on at the same time when we are working on a standard. We are finishing up work on the latest generation of the standard, while at the same time beginning the period of idea generation for the next version. And here I am talking about the main branch of 802.11 that results in the versions that are named and known to the public, such as 802.11n and 802.11ac.
Idea generation is a long process these days. We all discuss what we think is needed in the next version of the standard and what should change from the current version. The initial discussion typically occurs in the Wireless Next Generation (WNG) Standing Committee and/or in a Topic Interest Group (TIG). It then moves on to a Study Group, where the formal scope is documented in a Project Authorization Request (PAR). In these early discussions, there are often a very large number of ideas proposed. For example, during the recent discussions on the amendment being developed for the amendment after the next amendment to the standard,[4] there were hundreds of presentations over many months. Sometimes the ideas being presented are new and sometimes they are recycled from earlier versions of the standard or from ideas presented for earlier versions but not used in those versions.
Once we have all the ideas, we generally divide them into two buckets. There are the ones we know will not go into the next version of the standard and there are the ones we will contemplate for the next version. Not everything we contemplate will ultimately become part of the next version. We may reject or table some of the contemplated ideas during these discussions and the ones we think are worth putting into the next version will go through many changes and refinements before the end of the process and may still be rejected later on down the road. Sometimes when we amend a standard, we add new technology or features to the existing standard and sometimes we remove technology or features, and of course sometimes we do both at the same time.
Once we feel like we have a good initial set of ideas, we transition the collection of these ideas over to a Task Group that actually writes the amendment. A Task Group has regular meetings with agendas and established procedures for governance.[5] The Task Group follows Robert’s Rules of Order to try to make sure everyone gets an opportunity to be heard. Once the Task Group has developed a draft standard, the Working Group is responsible for the first stage of approving the draft via a formal voting, which we call balloting, process.
The IEEE Standards Association is a consensus organization and, for the IEEE-SA, “consensus” means that a ballot passes if it gets at least 75% support. If someone votes no, they have to give a reason, and each of those objections needs to resolved or answered before the process continues, even if the ballot has achieved 75%. During the first rounds of balloting, the draft often gets updated and changed significantly as people determine what’s been left out or what needs to be fixed or modified to make it work better. The group can stop accepting new ideas once a ballot reaches 75%, but sometimes they are accepted even that late into the process if it makes the next version better. Typically, the refinement process continues through multiple ballots until the approval vote is in the high nineties and nearly all of the comments have been resolved or answered to the satisfaction of the commenters.
Once the Working Group ballot is complete, the whole process repeats with a different group of voters that are intended to represent additional stakeholders, and not just the Working Group members. This is called an SA Ballot. All this balloting can take significant time (many months, often much more than a year) but the result is a better standard.
Most standards organizations have analogous processes to achieve a final proposal. While each method has different pro’s and con’s, in my view the IEEE-SA process provides an excellent balance between speed and quality, while giving all stakeholders a real say in the outcome. It certainly seems to have worked well in the case of the 802.11 standard, if one notes this standard supports over four billion new products per year with more than 17 billion WiFi devices in operation today.
That seems like a very involved process. How long does it usually take?
The process to develop each new amendment to the main branch takes 5-7 years. But because much of the work on different generations of the main branch occurs in parallel, it generally takes no more than 5 years for products implementing the next iteration of the standard to come to market. The process requires a massive investment by industry. There are currently about two hundred people in the 802.11 Working Group that meet face to face somewhere in the world every 2 months, and more often virtually, to work on it.
That’s a lot of people. Who are the people and how do they become members?
The IEEE-SA has both an individual and an entity based membership model. The IEEE 802.11 Working Group uses the individual membership model. This means that people are members, not companies. To get membership, a person needs to attend a certain number of meetings. Each member is supposed to vote their own minds, as an individual, and not as a company block. Of course, this does not work perfectly and these days people from certain companies seem to vote in blocks rather than individually. But, for the most part, people do vote based on their professional judgement.
This way of doing things is not the same in every standard development organization. For example, ETSI currently uses a model based on consensus between companies that seems to allow a small number of companies to veto ideas. For the formal voting process, ETSI’s model is “pay to play” with companies able to pay higher membership fees in exchange for more votes and influence. I have not been terribly impressed with this approach – it essentially excludes smaller companies that are less able to pay these high fees and gives undue power to those companies willing to pay a lot.
The Wi-Fi Alliance is different again. They use a two-level approval process. The first level is the task groups where the specifications are developed. In the task groups, companies are members and each company-member has an equal vote. So small implementers have equal standing with large companies in the task groups. But the Wi-Fi Alliance also has a Board to which Sponsor companies appoint Directors. The idea is that companies are free to vote for their own individual economic self interest at the task group level, but the Board’s Directors are supposed to be looking out for the entire industry and all the member companies interests as a whole. Again, the model is not perfect but most of the Wi-Fi Alliance Directors do a pretty good job of looking out for the industry’s best interests.
What happens behind the scenes when you’re working on a standard?
There are a lot of politics that go on behind the scenes, a lot of what I call hallway consensus building. People try to convince others to vote for their ideas. Sometimes a person agrees to support someone else’s idea in exchange for the other person’s support of a that person’s idea on a different topic.
Chip makers have a huge influence over the process because they have to implement the standard into their products first. They tend to send lots of people to meetings. They are the start of the value chain: the first line implementers. If the chipmakers are not willing to make it or think they cannot make it work or that it will be too expensive to implement, it will not happen. We can design great technology, but if the chipmakers cannot or will not implement that technology then that technology will not get implemented.
The process does not work perfectly but mostly it has worked well.
You’ve talked a bit about why the chipmakers participate in developing standards. In your experience, why motivates companies to have their people participate and contribute to standard development?
Companies participate and contribute when (i) they sell or intend to sell products that implement the standard; (ii) they have the technical expertise to participate and (iii) the feature set in the standard is important to them. This has historically been true of all the participants. Chipmakers participate because they are the first line implementers and they want to ensure that the feature set is one that is reasonable, cost effective and possible for them to implement.
Enterprise level companies contribute because, not only do they use those chips in their products, they tend to have a lot of experienced engineers and technical staff and the feature set – interoperability, power management, security, network management, quality of service – is important to them. It is also why consumer product companies have not historically been motived to contribute – they may use the standard in their products, but they do not have the technical expertise to meaningfully contribute and they do not care as much about the feature set or how the standard works, as long as it works.
In some standards bodies, for example 3GPP,[7] you also see a lot of service providers participating. These service providers rarely make their own products, but they could be considered to be the super users of some types of standardized products. They buy a lot of those products, some to use in their own networks and some to resell to their own customers. They serve as a proxy for what they perceive to be the needs of their own customers. They participate because they both (re)sell and use the products that implement the standard, they have the technical expertise to participate and the feature set is important to them.
Why did people start the initial development of what became Wi-Fi?
Networking and telecommunications inherently involve connecting to others. In telecommunications, things need to be interoperable in order for them to be useful. Back before there was the 802.11 standard, a few companies were trying to do wireless themselves. But it was not working very well to sell products that did not interoperate. Most of the companies that contributed to the initial development of Wi-Fi did so because they wanted to be able to sell more products. In order to sell more products, those products needed to work well together.
Standardization is an amazing process where a group of competitors get together to try to create something we can all use. When so many companies all contribute into a single standard, everyone has the benefit of everyone else’s ideas and contributions. Often, each idea is not individually that useful, but they are incredibly powerful when we put them together. The totality of all of those contributions, the standard, really is greater than the sum of the parts.
And, at the end of the day, Wi-Fi has been one of the most successful standards. It is projected that implementers will sell, collectively, over four billion Wi-Fi products next year. It is extremely unlikely that companies would have been so successful if they were selling proprietary devices that did not interoperate. So companies participate because, ultimately, by contributing to the success of the standard, they can implement that standard into their own products and contribute to their own success.
You have spoken a bit about implementation. And I know you were at Radiata which was the first company to sell a chip that implemented the 802.11a standard. What does it take to make a successful product that implements a telecommunications/networking standard?
Implementation is usually not an easy thing. As I said before, a standard is just a piece of paper. Turning the words on the page into a working chip can be incredibly difficult and take a lot of engineering time and expertise. At the end of the day, there often are wide variations in how well different chipmakers implement a standard into their chips.
Back in the early days of Wi-Fi, Atheros and Radiata were the two leading startup chip companies that were trying to implement 802.11a into their chips. Both companies managed to make 802.11a chips but their solutions worked differently. A number of more established chip companies also tried to make 802.11a chips in the early days but it proved quite hard to do and most of them were not successful.
The standard was not in and of itself sufficient to produce an effective and useful chip. Chip makers had to spend lots of time developing their own technology around the standard in order to make functional chips. You needed a lot of smart people to make workable chips even though there was a published standard at the time. Radiata was purchased early on by Cisco, mostly because of those smart people. Atheros become a key 802.11 chip vendor and much later on was acquired by Qualcomm. Although Qualcomm was already a successful cellular chip company at that time, I assume Qualcomm purchased Atheros because Qualcomm recognized that it would be easier to acquire the products and experienced people rather than develop the technology in house.
Also, at least in the telecommunications space, standards focus on things that you need to make products interoperate. But there is so much more stuff you need to make a successful product. My home has about 30 Wi-Fi devices floating around. These run the gamut from my laptop to my access point to my cell phone to my smart home devices (including even my bathroom scales). Each of these products uses chips that implement the 802.11/Wi-Fi standard, but each product is a very different type of device, with a wide variety of differentiating features. You need differentiating features to sell products. If everyone’s product is the same – if it implements the standard in the same way and does not have any differentiating features – then the only thing you can compete on is price. Most successful standardized products are successful because of their other features. Implementation of a standard is only a very small part of making a successful product.
You do not hear much about this, but successful implementation of a wireless standard also requires the right regulatory environment. One of the reasons for the success of 802.11 is that a number of countries opened up the 5 GHz band at about the same time. The development and implementation of 802.11 would not have succeeded without this band being opened up for wireless use. The same will be true going forward in the 6 GHz band if we can continue to persuade regulators to open up the 6 GHz band in order to significantly increase the socio-economic benefits of Wi-Fi.
[1] This is a summary of discussions which took place over several days. In order to ensure that my summary accurately reflects Andrew’s words and ideas, I gave him an opportunity to review it before posting.
[2] Some additional details from Marta: By 1990, a number of companies had wireless products that used proprietary technology and were not interoperable with each other. The original 802.11 standard development began in the IEEE in the late 1980s/early 1990s with the goal of creating a uniform wireless standard for interoperability. The 802.11 working group officially formed in 1990. You can find the early meeting minutes here: https://www.ieee802.org/11/Documents/DocumentArchives/.
The first version of the 802.11 standard was released in 1997 and is often called the 802.11-1997 standard. It was not successful – the developers had not specified enough information in the standard to make products implementing the 1997 version interoperate. The 802.11 working group went back to the drawing board and two years later, in 1999, the IEEE released two different versions of the standard. To be fair, one of those versions was largely based on the 802.11-1997 version of the standard but included additional features that allowed for greater interoperability. These versions are commonly known as 802.11a and 802.11b. Andrew’s involvement in developing and implementing the 802.11 standard began shortly after those versions were released.
[3] Some additional details from Andrew: Radiata was involved in one of the earliest amendments to the standard, often simply called 802.11a, which expanded the standard’s original scope from just the 2.4 GHz band into the 5GHz band. While 802.11a was not particularly successful in the market, it started the transition of Wi-Fi into the 5GHz band. The current main branch amendment is known as 802.11ax (aka Wi-Fi 6). The next amendment to the standard, 802.11be (aka Wi-Fi 7), is just around the corner. 802.11’s scope today includes what is generally known as Wi-Fi, but it also includes other technologies, including WiGig (wireless that operates in the 60GHz band, 802.11ad and 802.11ay) and HaLow (wireless that operates in the 900 MHz band, 802.11ah).
[4] That version is contemplated to be released after 802.11be. It currently is known as “Ultra Wideband Reliability” but will probably become 802.11bn when it is released.
[5] Over time, the processes and responsibilities have become more codified. 11-14-0629-10-0000-802-11-operations-manual.docx (live.com)
[6] The Wi-Fi Alliance is not a standard development organization but instead helps drive development and implementation of the Wi-Fi standard.
[7] The 3GPP (the 3rd Generation Partnership Project) is not a standard setting organization but instead develops cellular telecommunication specifications that are then transferred to standard setting organizations such as ETSI to be voted on and potentially adopted as standards.