Standards Best Practices Blog

We saw in my previous blog entry that standards in 1929 were generally used for reducing variety, for simplification, for reducing the numbers of items produced, eliminating the outliers, and thus reducing costs of manufacturing. This was just a step further in the industrialization of the U.S. and the mass-production movement that gave us interchangeable parts and standard clothing and shoe sizes.

That’s not to say that that was all that was being done; perhaps the writer of the Britannica article was simply caught up in the emphasis of that decade. There had certainly been a lot of standards efforts in interchangeable parts and manufacturing processes, and many of the “grandfather” organizations of today such as ASTM, IEEE, and ANSI had their roots in work started in the late 19th century.

But it’s now 2007. What does standardization look like today? It’s certainly more extensive; rather than the small handful of organizations developing standards back then, we have hundreds. And rather than focusing mostly on variety reducing standards we have standards in dozens of categories.

ISO defines a standard as a “document, established by consensus and approved by a recognized body, that provides, for common and repeated use, rules, guidelines or characteristics for activities or their results, aimed at the achievement of the optimum degree of order in a given context.” Or, as I generally paraphrase, an agreed-upon way of doing things to achieve a specific result.

The definition gives us a lot of wiggle room as far as the purpose of the standard, the means used to develop and approve it, its contents, the means defined by the standard to achieve the stated goal, etc. And this is appropriate: there is no “one size fits all” that will satisfy all conditions, requirements, technologies, etc. for all industries and audiences.

That leads to the creation of many different types of standards, or different ways to classify them. And by that I don’t mean just different types of technologies or different industries. Here’s just a few of the different classification schemes that I’ve seen while reading various descriptions of standards:

• De facto vs. de jure; whether the technology or product is simply accepted or widely used by the market, or it is a formal specification that has been approved by a recognized or accredited body.
• Mandated vs. voluntary; whether use of or compliance with the standard is required by law, or if its use is voluntary and optional.
• Anticipatory vs. retrospective, proactive vs. reactive; whether the standard reflects the development of a new, perhaps bleeding edge, technology, or if it is simply the codification of an existing technology or practice already in use by industry.
• Open vs. proprietary; whether the standard is freely available for all parties to use without the requirement of licensing or payment of royalties.
• National, regional, etc.; whether the standard is designed, approved, or mandated for use within specific geographical or political boundaries.
• Physical/measurement/unit/reference. These are the oldest standards, going back hundreds and even thousand of years; the biblical cubit, for example. The oldest standards organizations were created to define things such as railroad track gauge, screw threads, etc. This also includes the imperial and metric measurements for inch, foot, mile, pound, and meter, liter, etc., time measurements such as second, minute, hour, or the calendar, or the octane of gasoline, the Kw/hr for electricity, etc.
• Performance. These standards, sometimes government mandated, are for such things as vehicle miles per gallon or the efficacy of pharmaceuticals
• Safety standards, such as for the lamp cord that has an Underwriters Laboratory sticker on it, fire protection equipment, etc.
• Compatibility/interface/interchange, including screw threads and railroad track gauges (see above) but also all of the plugs and wire connections we use every day for our computers and other electrical appliances. But this isn’t restricted to physical objects; data models and syntax for information exchange at various levels in the stack all fit in this category.
• Variety reducing, such as clothing and shoe sizes, tire sizes, lumber, etc.

Note that this is not an exhaustive list; there are always going to be new ways to classify things. Note also that there is a great deal of overlap in these categories; for example screw threads fall into both measurement and compatibility, and tire sizes are both variety reducing and reference.

The short message, though, is that there’s a lot going on.

I’m browsing through my new (to me) Encyclopaedia Britannica, 14th edition, published in 1929. I picked this up for a very reasonable price from our local book sale, the Friends of the Tompkins County Public Library, where I volunteer. I really don’t know where I’m going to find the shelf space for these 24 volumes, but the illustrations are great. Plus seeing the snapshot of things as they were 80 years ago is pretty cool.

At random last night I picked up volume 21 (SORD to TEXT) and came upon the article titled “Standardization” and caught a glimpse of the world of standards from way back when.

Standardization is here defined as “…setting up standards by which extent, quantity, quality, value, performance, and service may be gauged. Instances are the miles, the hour, the pound, the bushel, and the dollar.” Britannica then quotes from Mechanical Engineering, August 1926, to enumerate the advantages to manufacturers, distributors, and consumers as follows:

“(1) Standardization stabilizes production and employment…

(2) It reduces selling cost…

(3) It enables buyer and seller to speak the same language…

(4) …it promotes fairness in competition…

(5) It lowers unit costs to the public by making mass production possible…

(6) … it makes deliveries quicker and prices lower.

(7) It decreases litigation…”

and etc. all the way down to item 16.

The emphasis of this list, and the rest of the Britannica article, is on reducing variety. A subheading at the end of the article is on Simplification, listing the advantages of “…the commercial elimination of unnecessary variety in sizes, dimensions, grades or qualities of common commodities.” A U.S. Department of Commerce study showed that 80% of business is done in 20% of the varieties in the product line (is this the first use of the 80/20 rule?), and cited the increased business profitability of focusing on the more popular lines and eliminating the rest.

Another interesting tidbit from the article was that the then-future President Herbert Hoover was a standards guy. “Early in 1921, Herbert Hoover, then president of the American Engineering Council (the executive body of the Federated American Engineering Societies), organized a committee of 17 well-known industrial and management engineers to make a survey of waste in the industry.” Hoover was originally a mining engineer, first for the USGS then as a consultant. After the first World War he headed humanitarian efforts in Europe, then later in 1921 was made Secretary of Commerce where he used his post to continue his theme of simplification.

On Friday of last week I was in Calgary (sans baggage; thanks United!) speaking at the Fifth International Conference on Standardization and Innovation in Information Technology, SIIT 2007, hosted by the University of Calgary. As I was in Washington DC for ANSI World Standards Week on Wednesday and Thursday I missed the first day (and apparently some good papers), but was there for the second. Slides for presentations may be found on the conference web site.

My presentation (paper and slides) was on the convergence and harmonization of accredited and non-accredited standards organizations. I introduced my presentation as being a bit like Mythbusters, but unfortunately without the explosions. There are lots of perceptions about consortia and about accredited or recognized standards bodies, but these perceptions depend a lot on who you ask. While many perceptions are based on historical practices these are changing and the two types of organizations are becoming more and more like each other. The issues of openness, democracy, balance, process, governance, expense, etc. are not binaries; each may be measured along a spectrum, and generalizations such as “all consortia are closed groups” or “all accredited organizations have slow processes” are simply not true. All organizations have the same business challenges of financing their work. Recognized bodies are developing processes that allow consortia work to be submitted for national and international approval, and consortia are participating in these programs because the recognize the value that such recognition can give.

I suggested further that many non-accredited standards organizations are not accredited simply because they don’t see the benefit in doing so, but are otherwise qualified to become accredited. Rather than looking at accreditation vs. non as a way to describe an organization, a better way to distinguish standards organizations would be to look at the organization’s emphasis on anticipatory vs. retrospective standards, the levels of satisfying industry and market needs, use of modern infrastructure, their openness to public review and inspection, emphasis on non-sales revenue, and the extent to which they promote adoption.

I’ll mention just one other presentation, the one given by Dr. Laura DeNardis of Yale University on the political implications of open standards. She says “Technologies embody values and, once developed, have political implications.” Standards can define the capabilities of, for example, an electronic balloting system or the availability and openness of public documents. Further, technologies and standards can define what is allowed by law when then they define what is technically possible. Laws and government activities in the areas of individual privacy, property rights, free speech, disaster response, national security, and global trade are all influenced by standards and what is technically possible.

DeNardis also discussed the concept and definition of openness in standards, and referred to work done by Ken Krechmer of the University of Colorado. Just as in my comparison of standards organizations, openness of standards is not a binary yes/no, but rather a spectrum of openness using a variety of criteria during development, implementation, and usage. The question is not how open something is, but whether the level of openness is appropriate for the specific context. I’ll recommend that you take a look at the slides for this presentation.

Last week I was in Washington DC attending ANSI’s World Standards Week. This annual WSW event is comprised of meetings of the various ANSI committees and member forums. A major theme this year has been that of consumer confidence related to the issue of product recalls, many of which are for items from China, including most recently toys. This is a big issue as we enter the holiday gift-buying season; parents want to be certain that the toys they buy for their children are safe. Representatives from the U.S. Consumer Product Safety Commission and from various consumer interest groups attended the ANSI meetings, and the representative of the Toy Industry Association was on the hot seat for many of the discussions.

The issue of toy safety, which has been in the news almost daily for the last couple of months, is one of standards. But it’s not a question of the need for standards – there’s already a lot of them in place – but of enforcement. Given the nature of the enforcement and the quantities involved, this becomes an issue of conformance testing and certification. The enforcers of the safety standards, which are codified into laws and regulations, are the laboratories which test products to ensure that they comply with applicable standards. Examples of each type of toy, and each manufacturing lot, need to be tested. That’s a lot of work.

But who is qualified to do the testing? Can the manufacturers do the testing themselves? With which testing laboratories can they contract to do the work? Or should the government do the testing?

Part of the discussion at the ANSI meetings was around this topic. The CPSC is a candidate for the body that would certify testing labs to do the work, but ANSI is suggesting that this is a role for which it is better suited, both for the sake of separating the regulation-setting and certification roles and because there is precedent for ANSI certifying conformance-testing labs in other areas. ANSI has listed its role in support of consumer confidence among its top priorities for the coming year.

The broader lesson here is that standards by themselves are only part of any solution. Whether the standard is for a consumer product or not, the standards developer must define a means by which the standard can be enforced, usually be defining tests for conformance. How do you know if a standard is being followed or not? And further, who is qualified and/or authorized to decide? These are all issues to be settled as part of the standards development process.

As Andy Updegrove points out today it’s payback time as SC34 is having trouble getting sufficient votes to pass any of their ballots, as the countries who joined specifically to vote on OOXML have now lost interest in further SC34 work. As JTC1 voting requires 50% of P members to approve, if there are a lot of P members in a subcommittee who don’t vote then nothing will ever get approved. This is just the situation I warned against a week or so ago.

To continue my thread of blog posts on the subject of standards best practices, how is such a situation solved? The first step is, as I’ve suggested in several posts, is to prevent stacking by requiring some sort of waiting period, by requiring participation over a period of time before granting voting rights. The second step covers the other end of the timeline: The committee process should include some provision to get rid of members who don’t participate.

There are, of course, a number of valid reasons why a particular member doesn’t participate in any particular ballot. The member may not feel that he has the technical expertise to form a valid opinion on the topic, or perhaps a member hasn’t had sufficient time to acquaint herself with the specification under review. Travel and other work activities, as well as holidays and vacations, interfere. The standards organization should provide sufficient time for a member to review the specification and form an opinion, but it should also provide members the opportunity to abstain from voting. The organization should never expect 100% participation by eligible voters, yet should require a sufficient number to ensure that the voice of the committee or other consensus body is being heard.

What should the organization do with a member who simply does not participate? Most organizations have rules regarding the retention of voting rights; these rules would generally mirror the rules for gaining voting rights. The organization could require, for example, that a new member attend two out of three meetings for some period of time before being able to vote, then would require the same level of meeting attendance plus the return of two out of any three ballots in order to retain voting rights. These rules should take into account valid reasons for non-participation, yet recognize that if a member is truly interested in participating he or she would do so.

But even with any fair set of rules it will take time to remove inactive members from the voting roster. Inactive members should be encouraged to resign if their presence on the voting roster is preventing other work to proceed. It is important to structure the committee process to encourage participation, provide for valid absences, but not penalize active members who want to get work done within the committee.

Standards organizations pay a great deal of attention to the development and approval of new standards. But if that’s all that they are doing they’re only doing part of the job. A standard that is not adopted, that has no implementations, is wasted effort.

The standards organization can do a number of things to promote adoption of the standard. These occur during all phases of the organization’s process and activities.

First, the specification that will become the standard must be designed with adoption in mind. The specification must solve a specific problem or meet a specific need. The committee or work group developing the specification should start by identifying those needs or problems through the creation of specific use cases. The specification must be well engineered to prevent ambiguities and undefined behaviours. The scope should be well defined to prevent feature creep or the inclusion of out-of-scope features, and the specification should reuse or reference existing work.

The committee should not be restricting its work to creation of the specification itself. Conformance requirements and/or a conformance test suite should be part of the committee’s work, as well as the creation of reference implementations or proofs of concept.

The organization’s process contributes to the quality of the specification as well. The development process should be open to all interested parties, and balance should be sought from a variety of interests and not just from vendors or manufacturers. Reviews, both public and committee, should be required, and the organization should provide facilities to support accurate and complete disposition of received comments. The process should require some number of implementations of the specification before approval to ensure that the specification is indeed implementable. Testing of the interoperability of those initial implementations will further prove the quality of the specification.

After development of the specification and its approval as a standard, the document must be distributed. This phase also has a great deal to do with the adoption of the standard. If the standard is not easily obtained then it is not as likely to be implemented. This includes making it easy for potential implementers to find the standard by publishing information about it on the organization’s web site and in databases of standards. The process for buying and/or downloading the document should be simple, and the cost of the document should be reasonable from the implementer’s perspective.

Even after publishing the standard, the organization’s work is not completed. The organization should engage in promotional activities such as sending out press releases, conduct educational workshops and seminars, host interoperability plug fests, and engage test labs to conduct third-party certification testing. And finally, the organization should have a plan for the maintenance of the standard, including ongoing collection of comments and then periodically revisiting the specification for updates and reapproval.

Keeping in mind the adoption of the standard throughout its development, approval, and after will go a long ways towards ensuring the success of the standard.