You are here  Software

Standardisation: Who's setting the standard?

Danny Bradbury
Friday 13 July 2001 01:05
As the need for computers to communicate with one another grows, so too does the need for universal standards. Easier said than done.

It is hard to believe in the 21st century that when travelling to a foreign country you still have to take an adapter so that you can plug your modem into the hotel telephone socket. You also have to take another adapter to plug your laptop computer into a power socket. For all their red tape, sombre debating sessions and paperwork, standards bodies still have a lot to do when it comes to interoperability.

The IT sector makes other industries look streamlined and efficient. Computers are complex and the suppliers will not let customers forget it. The industry is littered with jargon and acronyms, many of which refer to particular standards implemented to solve specific computing problems.

It is odd that since the emergence of the open computing movement in the early 1990s things seem to have got worse. Then, individual suppliers had their own standards (such as IBM's systems network architecture), but there seemed to be fewer of them. Customers signed up with one supplier and used everything they had, so there were no cross-platform issues. Now, there is a standard for everything from personal information management software synchronisation to digital resource management.

"Are standards all about connectivity?" asks Peter Jones, president of software integration company Mercator and vice-president of the Computing Services and Software Association (CSSA), a trade body for software and services companies. It is a valid question. One reason why standards have proliferated is the increase in connectivity in an Internet-enabled world. As the need for computers to talk to each other continues to grow, the demand for a lingua franca increases proportionately.

The problem is that you cannot have a single communication language covering a complex set of functions. This is why, when the need for standards became apparent, the International Standards Organisation's Open Systems Interconnectivity (OSI) stack became so useful. The OSI stack was a way of categorising different levels of functionality into layers, from a low to a high level of operation.

At the bottom (layer one) was the physical transport layer - the physical link used to connect one computer to another. The system went up through several layers, each one reliant on the one before it, until it reached the application layer at the top, which used the technologies in the layers underneath it to accomplish application-specific tasks.

Layer seven - the top - contained its own standards, such as the X.400 standard for e-mail delivery. It is impossible to cover the OSI in depth here, but it is an important industry development. For more information see John Larmouth's excellent guide to OSI at www.isi.salford.ac.uk//books/osi/all.html#head1 .

Twenty years ago if you looked at any layer in the OSI stack you would have seen different technologies on offer for corporate use. Each supplier had its own way of tackling a problem, which is why IBM came out with Token Ring, which was eventually adopted by the IEEE consortium. The standards landscape changed dramatically when TCP/IP came to the fore.

Originally invented by the US Defense Advanced Research Projects Agency as the basis for the Internet project, TCP/IP grew in popularity, due in no small part to the efforts of the early Unix community which promoted it as part of the system.

Many people assume that TCP/IP is simply the transmission protocol for the Internet, but in fact it is a whole stack on its own. The TCP/IP stack, which became prominent at the beginning of the 1990s, has all but replaced the OSI stack in many areas as a framework for different protocols, with standards such as the simple message transfer protocol, Internet protocol (IP) and others becoming dominant.

The adoption of TCP/IP on a universal basis was vital if the Internet was to grow, and Microsoft's inclusion of a TCP/IP stack as an out-of-the-box component of the Windows client in 1995 sealed the deal; TCP/IP became one of the most important sets of standards in the past decade.

So, with the world centred on the Internet, why are things so confused in the standards sector? Chris Gabriel, regional manager at Enterasys Networks, thinks he knows. Enterasys is involved in defining many networking standards, including the forthcoming 10-Gigabit Ethernet standard. He believes that at lower levels down the stack, simple interoperability was the key. Few stakeholders cared about low-level standards such as Ethernet, as long as they worked and meant that different suppliers' equipment could work together. As you move further up the stack, functionality becomes more complex and each stakeholder has a more vested interest in what goes into a standard.

Peter Sharpe is chief scientist at extensible markup language (XML) editor software company SoftQuad and was a co-founder of the XML standard, thrashing out the protocol with a team of nine other experts on SGML (XML's predecessor). He draws a distinct parallel between the definition of the original XML standard and the XML Schema standard, recently passed as a recommendation by the World-Wide Web Consortium (W3C) after a long period of political wrangling.

"XML was in its first working draft within four months," Sharpe explains. "It took in total a year and a half, but most of that was working out details. XML Schema took two years and it took a long time for even the first draft to come out."

He says that most work on XML was done without interference from the W3C due to a lack of interest in the language at the time. The small committee created a 30-page standard, which was relatively simple as software standards go. By the time XML Schema came up for discussion, everyone wanted a say.

"The Schema group was a huge committee - there were 40 people on it and everyone had their own agenda," Sharpe says. "There were big companies there, with IBM, Oracle et al pushing hard with their stuff."

The origin of a standard will affect the speed of its development and its range. Standards come from a variety of areas, including industry consortia, government-sponsored bodies such as the United Nations and even individual companies. But you can subdivide these sources further, according to Brian Whetten, chief scientist at middleware company Talarian. Whetten is heavily involved in the definition of standards within the Internet Engineering Task Force (IETF), but has also worked on the Java Messaging System technology that is at the heart of Sun Microsystems' J2EE server-side Java suite.

Different standards definition bodies have different cultures that can affect the types of standard they consider and the way they develop them, says Whetten. At the IETF, the majority of people are on some company's payroll, but they come from an academic background (75% of IETF members have a PhD). Such qualifications make those people suitable to define fundamental standards that would sit at a low level in the stack. These technologies need to be very robust, well thought out and comprehensive, which is what the academics are good at, but they often face huge deployment challenges. The definition of IP version 6 is a prime example; it needs a very thorough, academic mindset to address the protocol definition that is so fundamental to the Internet future, these people will focus only on technology, leaving its transfer into the commercial world to others.

Whetten has seen an evolution in the way standards are defined. In the past, standards have been defined using this academic approach, on a "bottom-up" basis, starting with low-level technologies and building up requirements until they reach cosmetic areas such as the programmable interface. This process has worked well in the past, because academics were pioneers at the frontier of the Internet, developing technologies and standards to be implemented in new territories. (Before IP was developed there were commercial systems with proprietary networking standards, but no Internet. Hence, there was effectively no legacy.)

Now, with a plethora of existing systems and platforms so well established, such fundamental, non-commercial standards definitions become less appropriate, argues Whetten. Instead, a "top-down" approach is needed, where business requirements are defined at the start and the standard definition process is more supplier-driven.

Such a business-centric model is not without challenges. "Top-down development has a political risk - you can get separate camps," Whetten says. "If Microsoft decides that it has a standard, and Sun has one, you end up with these islands of connectivity. That's the real challenge - how do you breach the top-down standards that create these islands?"

This is what makes XML an interesting standard. It was developed by a technical committee with a strong academic background in a bottom-up scenario but, says Whetten, the business-centric, industry-specific document type definitions were developed in a top-down model. Hence, the XML standard is unassailable and easy to follow, but the complex fabric of vertical market XML-based languages is confusing and has possibly hindered the growth of XML-based integration.

One good thing has come out of the top-down approach - increased opportunity for customer input. Although different committees and industry bodies vary in attitudes towards end-users, the companies participating in a given standard definition process are accountable, says Whetten. You can think of it as a parliamentary system, in which the companies are the MPs and the customers are their constituents.

"Certainly, if a company has a large customer that is demanding something that could be achieved through a standard, then that company's representative on the committee will push very hard," Whetten explains. "There is lobbying at each level."

This contrasts sharply with the role of government in computing standards definitions. Government does not have the knowledge base to legislate on standards, says Jones. "Because it doesn't, it ends up trying to recommend an answer that doesn't get support from suppliers or customers in the industry."

The most streamlined standards definition process of all involves the creation of a technology by a single company with enough power to push it through or with the luck or intelligence to introduce a technology just when it is needed. Good examples are Microsoft, which has created numerous standards including everything from open database connectivity through Soap and even arguably to Windows itself; the Win32 API may be proprietary, but unless you code your programs to comply with it, sales will suffer.

Another example is Sun Microsystems, which introduced Java and then hyped it up as a necessary technology for the network computing revolution, which fell flat on its face in the mid-1990s. Nevertheless, Sun made Java stick by switching the emphasis to the server and touting it as a platform-independent e-commerce development environment.

But for all of its "openness" with Java, Sun is still selling a proprietary technology. It has made it popular by making licences available to everyone, while failing to hand the technology over to standards bodies such as ISO or the European Computer Manufacturers' Association. No matter; the company has managed to get the best of both worlds, creating technology from which it can still make a profit, while simultaneously promoting its openness through a technology definition process (the Java community process) that is open to input from other companies. Such a strategy is the ultimate logical conclusion of the top-down approach - if you cannot get the industry to do what you want to do, then become a for-profit standards body yourself.

Jones sums things up nicely when he talks about what drives standards. Standards become vital when the cost of running systems in a fragmented market exceeds the cost of developing and implementing a standard. Now, with companies being encouraged to increase internal efficiency and to do business with each other, adherence to standards is increasingly important.
<br>Industry is inefficient when it comes to defining standards for hardware and software interoperability, it is not customers that are slow to take up standards. Some standards initiatives have worked well and produced admirable results, but others (typically those defined in a top-down process, as characterised by Whetten) have generated confusion and frustration. In a world where complexity is a constant, this is the last thing customers need.


HOW TO BE A STANDARDS POWERHOUSE
Few customers have the time or the resources to spend on standards definitions. They require large amounts of employee time for meetings and paperwork and unless you are an IT firm, it does not make good business sense.

The best way around this problem is to cement a relationship with a supplier whom you know to be politically powerful. Generally money talks and if you spend a sufficient amount with one company it will fight your corner in the committee room. Fostering personal relationships with key technical staff can do you no harm and highlights the similarity between the parliamentary process and the IT standards world.

Alternatively, if the technology standard is sufficiently important to your company, find out if there is an industry-led initiative. A good example of this is the Global Commerce Initiative, formed in 1999, which facilitates end-user input on global electronic trading standards to the major standards bodies.

The other option, if you are powerful enough, is to start up your own vertical market initiative. This could take the form of a standards definition process or could involve the implementation of a business-to-business trading hub, which will give you the opportunity to pull in trading partners and suppliers, persuading them to make use of your facility. This provides you with a platform from which to define trading policy and standards in the future.


STANDARDS TIMELINE
IT standards have come a long way since the
industry first realised that computers had to work together. It is impossible to document all computing standards and bodies without taking up a whole book, but here are some milestones:

1865
International Telegraph Union (now ITU) formed (www.itu.int)
1947
International Standards Organisation launched (www.iso.ch)
1974
IBM introduces systems network architecture
1983
US Department of Defense standardises on TCP/IP
1986
Internet Engineering Task Force formed (ww.ietf.org)
1988
The European Standard Telecommunications Institute (ETSI) formed (www.etsi.org)
1991
European nuclear research laboratory CERN releases the World Wide Web
1992
Internet Society formed, providing legal framework for IETF
(www.isoc.org)
1994
World-Wide Web Consortium formed (www.w3.org)
1995
Version 6.0 of the Internet Protocol released
1998
XML 1.0 becomes a W3C recommendation
2001
ebXML e-business exchange standard ratified.


A SOFTWARE STANDARDS WISHLIST
In spite of the proliferation of different standards in the IT sector, there are still areas that could do with some universal protocols to unite the market. This is what the bureaucrats have not yet covered:

Software quality
One of the things holding back the market for third-party, component-based software and the new Web service frameworks is a quality assurance standard. Many companies may feel uneasy using third-party software components unless there is some guarantee that they will work under specific conditions without crashing. To our knowledge, no such rubber stamp for software quality exists

Virus standard
Wouldn't it be great if virus writers produced viruses, worms and trojan horses that operated along specific lines, as defined in the Standard Template for Unsolicited Programming Intrusions and Destructive Generic International Trojans specification? At least that way anti-virus companies could deal with them more easily. But that would never happen - most virus writers wouldn't even be able to spell it, let alone follow it

SpamML
Similarly, it would be useful to have a standard for unsolicited commercial e-mail (spam) in an XML format supported by all the major e-mail clients. Spam could contain an e-mail header with information about the subject matter. An XML-based syntax may help in the creation of an "opt-in" system for communicating legitimate marketing e-mail to end-users

Instant messaging standard
The world needs an instant messaging standard. At present, with messaging systems from companies including AOL, Microsoft and Yahoo, people often know friends, workmates and family signed up to different services, meaning they must run all the clients at once. A single interoperable standard would make things easier for the end-user. The IMUnified initiative has produced an interoperability proposal, but AOL is not a member. It plans to introduce its own proposal this summer, while the IETF is also work on an interoperability document

Meta standard

A standard for standards development is the most necessary standard of all, given the disingenuous nature of today's many different standards definition activities. If the industry could agree on a way to develop standards that every supplier stuck to without confusing end-users and managed to avoid replicating its own efforts with multiple rival standards, the computing industry would be a happier one.