How VRML became an ISO Standard

Dateline: 5/25/98

The following article is reprinted in its entirety by permission of the authors and represents a definitive history of the way in which VRML moved from an Intenet community effort to a formal ISO standard.

The Development of the VRML 97 International Standard

Authors: Rikk Carey, George S. Carson and Richard F. Puk

(c) Copyright Rikk Carey, George S. Carson and Richard F. Puk 1997. Permission to copy and distribute this document is hereby granted provided that this notice is retained on all copies and that the document is copied and/or distributed in its entirety without alteration. This work has been submitted for publication. Copyright may be transferred without further notice and the accepted version may then be posted by the publisher.

Introduction

International Standards in the area of Information Technology are increasingly being developed by standards committees in partnership with organizations such as industry consortia. A good illustration of this is ISO/IEC DIS 14772-1, the Virtual Reality Modeling Language (VRML) Standard, which has been standardized by Joint Technical Committee 1 (JTC 1) of the International Organization for Standardization (ISO) and the International Electrotechnical Commission (IEC) in partnership with the VRML Consortium.

Of as much interest as the technical content of the VRML standard is the methodology used in its development, which points a way forward to developing future information technology standards in a similar manner. The early stages of development were characterized by an open Request for Proposals, debate over the relative merits of the submissions using an Internet e-mail discussion list, and selection of a winner. The final stages of development were carried out as a cooperative effort between ISO/IEC JTC 1 and the VRML community. Specific techniques were used to avoid problems that have caused some other recent standards efforts to fail.

The VRML 97 specification is now Draft International Standard (DIS) 14772-1 and is currently being balloted at this level within ISO/IEC JTC 1. The DIS text, like the CD text before it (the "VRML 2.0 Specification" that was widely distributed in August 1996) was cooperatively developed by ISO/IEC JTC 1 and the VRML community. The DIS ballot closes later this year and the final International Standard (IS) text will be published by both the VRML Consortium and ISO/IEC by the end of 1997. Thanks to the provisions of the Cooperative Agreement governing the joint development work, the final IS text will be available for purchase through ISO and IEC but will also be freely available over the World Wide Web (WWW) from the Consortium’s web site (www.vrml.org).

Why is VRML Important?

The Internet, and especially the WWW, is changing how information is disseminated throughout the world. Hyper Text Markup Language (HTML,) the basis for the current WWW, provides a document-based interface that can incorporate raster graphics images (as GIF or JPEG files.) While the document paradigm is suitable for some forms of interaction, it has serious limitations both in terms of its information structuring and its interaction capabilities. VRML enables the next level of interaction, by moving the web beyond the document-oriented paradigm into virtual worlds based on 3D interactive computer graphics. A reasonable analogy of the capabilities of VRML to those of HTML is the comparison of the interactive experience reading a book to that playing a video game. The applications of VRML are broad, ranging from business graphics, to entertaining web page graphics, to manufacturing, scientific, entertainment, and educational applications, and of course to 3D shared virtual worlds and communities.

VRML is patterned on the success of HTML, which provides limited graphical interaction in two dimensions. In moving beyond the limitations of HTML, VRML has accomplished several things. First, it takes graphics from 2D raster images to 3D geometric (and raster) graphics suitable for both 2D and 3D interaction. While 2D graphical images are useful, they do not facilitate the investigation of models and other phenomena which are part of our three-dimensional perception. Second, VRML has been designed to be a file format capable of not only describing the shape of three-dimensional information, but also describing the interactive behaviors to be applied when a user encounters information and attempts to interact with it. It is this combination of behavioral definition coupled with a three-dimensional description of information which embodies the new capabilities introduced by VRML.

History of VRML Standardization

The Early Days

Several independent projects provided the foundations for VRML. In 1994, Mark Pesce and Tony Parisi became interested in the idea of a 3D user interface for the Internet. They developed an early prototype, called Labyrinth, using a donated copy of Reality Lab, the 3D rendering technology from RenderMorphics in the UK, and gave a talk at first World Wide Web Conference in Geneva outlining their vision. Independently of the work by Pesce and Parisi, the Open Inventor project was started by Rikk Carey and Paul Strauss at Silicon Graphics, Inc. (SGI) in 1989. The goal of this project was to build a 3D computer graphics environment that enabled a broad, multi-platform application framework to expand the role of 3D graphics into mainstream markets. Open Inventor included a file format designed to be a standard interchange format for interactive 3D applications. Pesce and Parisi approached SGI through various avenues to engage them in the effort to develop a 3D user interface for the Internet. The collaboration between the Open Inventor architects, Parisi and Pesce in producing VRML 1.0 was the result. Besides Open Inventor and Labyrinth, other projects that had influence on VRML were:

Annotator: A 3D annotation application developed by SGI's Inventor team. It was the first commercial 3D hyperlinking application.

OOGL: The University of Minnesota's geometry language.

CDK: Autodesk's Cyberspace Developer’s Kit

World Chat: World Inc.'s 3D Multi-User Dungeon (MUD) application.

Reality Lab: A real-time 3D software rendering library developed by RenderMorphics Ltd.

Following the talk by Parisi and Pesce at the first World Wide Web Conference, the www-vrml e-mail discussion list was formed. It was moderated by Brian Behlendorf and Mark Pesce. The list was devoted to developing the idea of a 3D interface for the Internet. After much discussion, the group developed a series of requirements and issued an open Request for Proposals for technology submissions. Five submissions were received:

Open Inventor's file format (SGI);

OOGL (University of Minnesota);

CDK (Autodesk);

Manchester Scene Description Language (Manchester University); and

Labyrinth (Tony Parisi and Mark Pesce).

After some discussion on the list, the proposals were voted on by participants. Voting was conducted by sending e-mail to the list moderators. The whole process was ad-hoc and informal, but had the advantage of being simple to implement. There were only a few hundred votes total. Open Inventor's file format received the largest number of votes by only a narrow margin, OOGL came in second and the other three proposals each received far fewer votes.

Thus, the Open Inventor proposal became the working document for the first VRML specification. Gavin Bell, Mark Pesce and Tony Parisi (with significant contributions by Rikk Carey and others) refined the proposal into a first draft that was presented at the second International World Wide Web Conference in October 1994 in Chicago.

The early appearance of VRML products and an early buy-in to VRML by major corporations were essential economic factors that supported the standardization effort. In April of 1995, SGI released WebSpace Navigator, the first VRML browser. SGI licensed WebSpace to Template Graphics Software to be ported to variety of platforms, including Windows. Intervista's WorldView 1.0 and Paper's WebFX were two other influential early browsers. Microsoft licensed WorldView 1.0 from Intervista and bundled it into Internet Explorer 2.0 in June of 1995. In February of 1996, Netscape bought Paper, integrated WebFX (later to be called Live3D) and endorsed the Moving Worlds effort.

E-mail discussion lists

Since the use of moderated discussions by e-mail played an important role in the evolution of the VRML specification, it is worthwhile to examine how these discussions actually worked in more detail. The key elements were:

It was useful that a list was moderated. The moderator(s) watched the discussion and insured that it stayed on track. It was rarely necessary for the moderators to step in to redirect the discussion since the list tended to be self-policing. If the discussion strayed or "flaming" began, other list members would usually steer the discussion back on track or calm down the enflamed participants.

In cases where there was a difference of opinion on an issue that could not be resolved by discussion on the list, one of the primary specification authors would typically drive the discussion to consensus. They would do this by making a proposal and suggesting that everyone support it for the sake of progress. They would remind everyone that every decision would have detractors. There were major disagreements on only a few of issues. Once a resolution became fairly clearly popular, most list participants would tend to accept the decision but might remind the list of their own opinions later if the topic was reintroduced.

Several thousand different people were on the www-vrml list, although less than a hundred people were typically active at any one time. Certain people would be active for a three to six month period and then disappear. Others have been active since the beginning of the list.

Today the list has about fifty old-timers that have been active (on and off) for a several years. Many of the regular posters from 1995-96 are gone. Also, there are about fifty recent posters whose focus is more content and user related.

The nature of the discussion on the list has evolved over time, shifting from purely technical and philosophical discussion to a mix of technical, philosophy, content or creative related, products, and newcomers seeking information.

The group dynamics of a large e-mail discussion list are certainly different from those of a small standards committee that meets face to face. To a large extent, the organizers and "old-timers" on a list can set the tone for it. In the case of the www-vrml list, these people were generally very helpful, gracious, and courteous to newcomers or strangers. Making a first posting to such a list can be intimidating. It is very similar to taking the microphone a conference with a thousand people in the room! If the newcomer was humble and polite and could support their position with good, solid reasoning and supporting arguments they were welcomed to the list. On the other hand, those who took an aggressive posture were often not welcomed.

Early work in ISO

In parallel with the evolution of VRML in the WWW community, there were several attempts to initiate a 3D graphical metafile project within ISO/IEC JTC 1. Prior to 1994, there was interest within the standards community, but for many years there was insufficient industry support to initiate a project. In 1994, JTC 1/SC24 issued a call to industry for submissions of candidate 3D formats. Presentations were made by SGI, Apple Computer and others at a meeting in Airlie, Virginia, USA in October 1994. Each of the submissions was carefully reviewed and compared against a list of requirements drawn up by SC 24. At the completion of this review, it was determined that the SGI submission based on the Open Inventor format was more general and more in-line with the requirements of SC24. It was also determined there was sufficient industry support for a New Work Item Proposal to be prepared and submitted to JTC 1 for approval. At this point, although the Open Inventor format had been selected as the base specification by both JTC 1 and the VRML community, the work in each arena was still proceeding independently.

The VAG is formed

During the summer of 1995, it became evident to the ad hoc organizers of VRML (Mark Pesce, Rikk Carey, Gavin Bell, and Tony Parisi) that more order was needed to better manage the evolution of the specification. Two things led them to this conclusion:

Many companies and individuals were calling for (even demanding) that new features be added to the specification. It became obvious that some of these companies would create their own "standard" if progress could not continue in an orderly fashion. Several companies also asked for more order as a condition for their continued participation (and/or non-competition.) These companies wanted a reliable and professional entity that they could work with, not just an informal e-mail discussion list. The ad hoc organizers felt that an (un)official body could provide the needed structure, acting as contact point, arbitrator, and technical leader for VRML, listening, distilling, and directing the future of VRML.

By this time, the VRML community had grown substantially, both in terms of active members and bystanders. The old model for discussion and consensus formation worked fine when the community was smaller, but became less effective as the group got larger. It became apparent that some entity needed to more actively manage the process and make decisions.

The exact form of the entity that was to provide the needed structure was subject to some debate among the organizers. Some felt that it was time to form a "VRML Consortium." This view was supported by several large, influential corporations that wanted to participate in the future evolution of VRML. Others were opposed because they felt that the VRML community still needed to move rapidly at this stage in its evolution, and that a consortium of competing companies would slow the process down. The perception that rapid progress was required was fueled by the realization that VRML 1.0 was inadequate and that a follow-on specification must be developed very quickly. There were legitimate concerns based upon witnessing the recent history of multi-company consortiums, that a large, formal organization would slow things down. The organizers eventually agreed to form a small, technical body to govern the near-term future of VRML, with plans create a proper, slow, bureaucratic consortium once the follow-on specification was stable enough.

Thus, the VRML Architecture Group (VAG) was formed. The goal of the group was to continue the technical evolution of the specification according to the consensus-based requirements of the VRML community. Besides Mark, Rikk, Gavin, and Tony, six others were invited to balance representation and provide expertise in other areas that VRML covered (e.g., networking, content creation, and multimedia). The other initial members were Brian Blau, then at Autodesk, Jan Hardenbergh, then at Oki, Jon Marbry of Microsoft, Tom Meyer, then a student at Brown University, Mitra, then at Worlds Inc., and William Martens. At this point, the plan was to start from VRML 1.0 and use the VAG as the design team that did the technical work necessary to create the follow-on specification with review of the VAG’s work on the www-vrml list. There were no plans to issue a Request For Proposals for a new specification.

ISO and IEC Get Involved

By late 1995, it was clear to JTC 1/SC 24 that support for VRML was increasing within the Internet community and all major vendors were supporting the specification. SC 24 realized that a largely separate ISO/IEC effort to develop a 3D metafile format was not justified (even if the ISO standard was "better" than VRML in some ways.) Thus the first discussions on merging the VRML and JTC 1 3D metafile work took place in late 1995 between the VAG and SC 24. Also by this time it was clear to the members of the VAG that even more formality and procedures were needed to manage the evolution of the VRML specification, and the long experience of JTC 1 in standards development could be beneficial.

At the same time as cooperation between the VAG and JTC 1 was beginning, there was a raging debate about the future of VRML on the www-vrml email list. Many contributors wanted a dramatic revision of the specification - one that included "behaviors" (a term that meant different things to different people). In the first week of December 1995, both SGI and Microsoft announced proposals for the future of the VRML standard. SGI's was called Moving Worlds and was based on a collaborative effort with Mitra and Sony Research. It was a significant advance over VRML 1.0, in terms of behaviors, interaction, and animation, but retained the VRML 1.0 (i.e., Inventor) philosophy and style. Microsoft's ActiveVRML was a dramatic departure from VRML 1.0. It was based on the ML programming language and provided a more general programming language solution to the problem.

During this period, the VAG moderated discussions on the list and guided the process. In the pre-VAG days as VRML 1.0 was finalized, this process had consisted of releasing various revisions of the specification and managing the comment resolution process. It very quickly became evident to the VAG that they could not create the VRML 2.0 specification on their own, so the group shifted gears and issued an RFP. From December 1995 on, they stopped acting as a technical committee and started acting as a process committee.

There are several reasons why the VAG was never able to function effectively as a technical committee. The lessons from the VAG experience are useful for other communities that might be considering forming an technical committee to create a standard. Some of these lessons are:

forming an effective technical design team takes years of practice as well as good luck;

ten individuals is too large a size for a design team;

the VAG had no leader (and most members did not want one);

the members had many diverse goals;

the members had different business objectives; and

the team members did not have sufficient experience.

In retrospect, the only reason that the VRML community was able to move as fast as they did adopting VRML 1.0, was because it was already designed, battle-tested, and mature from the Open Inventor experience. Thus, it was a comparatively simple design task to update and revise VRML 1.0. When the VAG was formed and it tried a more formidable design task (VRML 2.0) the problems as described above became obvious and eventually caused the approach to fail. Once the VAG transformed itself into a "process" committee, things improved. The task of managing the work was much simpler than doing it.

In mid-December 1995, the VAG decided to issue a request-for-proposals for VRML 2.0. The RFP was issued on January 4, 1996 with a deadline of February 2. The RFP asked for complete proposals and provided technical guidelines representing the requirements. Six proposals were received:

Active VRML from Microsoft,

Dynamic Worlds from GMD and others,

HoloWeb from Sun,

Moving Worlds from Silicon Graphics and others,

Out of this World from Apple, and

Reactive Virtual Environment from IBM Japan.

A VAG meeting took place on February 5-6, 1996 in San Francisco to draft the process for selecting the winning submission that would become VRML 2.0. At this meeting, representatives from ISO/IEC JTC 1/SC24 formally met with the full VAG for the first time and it was mutually agreed to pursue ISO standardization of the VRML 2.0 Specification that would emerge from the selection process. Immediately after the meeting, the VAG issued a statement that described a 6-week public review and voting process. Anyone was free to visit a VAG-hosted Web page, vote on each of the proposals ("Strongly in favor" to "Strongly against") individually, and pick their favorite proposal. To encourage discussion and debate, the results of the voting were available even during the vote, and each person could change their vote at any time during the voting period. The only restriction placed on voting was a stern admonition to read the proposals and vote based on the proposals' technical merits. Ballot-box stuffing was discouraged by requiring a valid email address (a vote confirmation message was sent to the email address given, and if the email was not deliverable then the vote was discarded), and by making it clear that the list of email addresses of everybody who voted would be made public. This voting process worked surprisingly well; just under 200 votes were received, which probably accurately reflects the number of people who took the time needed to do a fair technical evaluation of the various proposals.

The final polling results showed Moving Worlds a strong favorite, and VAG members unanimously agreed to recommend it as the working draft for VRML 2.0. At this point, ISO work began as the Moving Worlds specification was circulated within the SC 24/ WG 6 Metafiles Rapporteur Group as a Working Draft.

Formal Cooperation Begins

In June 1996, a representative of the VAG attended the JTC 1/ SC 24 meeting in Kyoto, Japan. At that meeting, the final details of a Cooperative Agreement governing work on VRML 2.0 were decided. The purpose of this written agreement was to govern the terms under which a public specification (i.e., one whose text is available for use by others with minimal or no restrictions) such as VRML was to be transposed into an International Standard. There were several key things that both the management of JTC 1/ SC 24 and the VAG wanted to achieve with this agreement:

Establish the VRML 2.0 Specification as the base document for the ISO/IEC work. This meant that all changes were to be "deltas" from VRML 2.0. Hence, the work was not to start from a blank sheet of paper.

Allow no "unnecessary changes" to the specification. What has happened too often in the past is that an ISO or IEC committee has acquired a base document from an outside source but the document has been heavily modified as a result of comment and ballot resolution processes. Changes were often made simply because someone claimed to have a "better" way of doing a feature, or to force change for the sake of needless compatibility with obsolete standards or those that are not widely implemented. Such changes have caused needless delays in standards processing and resulted in a standard that do not match commercial practice.

Establish lines of communication and liaison. For this purpose, the Category C liaison mechanism of JTC 1 was used. This type of liaison allows an outside organization to participate in a limited set of the program of work of JTC 1. To implement this provision, the VAG was appointed a Category C liaison to JTC 1/ SC 24.

Determine access policy for all documents. ISO and IEC have strict rules regarding availability of both working documents and final International Standards. Some members of ISO and IEC derive substantial revenue by selling access to working documents to companies and other organizations within their countries. Further, ISO and IEC both enjoy income from the sales of International Standards at prices that seem exorbitant to most in the information technology community. The Internet community on the other hand has always had a policy of free and unlimited access to all of its documents over the Internet. Since this was to be a joint development project, the Cooperative Agreement allowed each community to treat all project documents, including draft and final International Standards according to their own practices. This meant that ISO and IEC could sell the standards while the VRML community could make the same documents freely available at their WWW sites.

Agree on common text. Both sides desired that there be only one VRML Standard. The Cooperative Agreement specified that this identical, common text would be distributed by each organization at each key stage of processing. There would never be a separate "ISO/IEC Version" and a "VRML Community Version."

Find an acceptable document format. Tradition within the VRML community called for publication of the standard as an HTML document for on-line access. This had never been attempted before with any International Standard within either ISO or IEC. Special permission was granted by the Information Technology Task Force (ITTF) at ISO Central Secretariat in Geneva to allow VRML to be published in HTML as a trial project.

Move ahead with all deliberate speed. Both sides perceived a limited window of market opportunity for VRML. Many ISO and IEC standards committees have a well deserved reputation for taking an extensive amount of time to finalize standards, often releasing them too late to be used effectively. One reason for this is the attitude that standards from ISO and IEC must be "perfect". This is in contrast the prevailing attitude of the Internet community that a standard must be "good enough." Recent improvements in JTC 1 processes presented the opportunity for finishing the VRML standard within an 18 month window. This meant both sides accepting keeping to the schedule as being as important as the technical quality of the standard.

With the Cooperative Agreement negotiated and approved by both sides, joint work began on the International Standard version of VRML. Based on comments received on the Working Draft circulation, VAG and SC 24 representatives agreed on the changes that were necessary to create a document of sufficient quality to serve as an ISO/IEC Committee Draft. Two co-editors were appointed, one from SC 24 (Richard Puk) and one from the VAG (Rikk Carey) and work began to cooperatively produce the single document that would serve as the VAG’s VRML 2.0 Specification and as ISO/IEC’s CD 14772-1.

The Committee Draft Stage

It is useful to look at the changes necessary to convert the Moving Worlds Specification into CD 14772-1, VRML 2.0. These changes are typical of those that are made to specifications that come into ISO and IEC from industry. Such specifications are often "after the fact" descriptions of a single implementation rather than specifications purposely written to permit multiple, independent, interoperable implementations. Thus, writers of future specifications can learn from the experience converting the Moving World’s specification into CD 14772-1. The major changes were:

Organization of the specification. International Standards are all organized according to a common format developed over the course of many years. This format includes unnumbered Foreword and Introduction clauses; followed by numbered clauses such as Scope, Normative References, and Glossary; and then continued with as many standard-specific clauses as are required. Reorganizing the Moving Worlds Specification according to this arrangement was mostly straightforward, with the few exceptions noted below.

Consolidation of conformance statements. It has been found very useful if there is a single place in a standard where an implementer can look to find which features must be implemented and which are optional. This also applied to features that have an implied range of possibilities (such as the number of nodes that an implementation must support.) In the Moving Worlds Specification, such conformance information was scattered throughout the document. Consolidating it in a single Conformance Clause took a substantial amount of work. In the process, missing conformance information was also identified and added.

Use of International English. International Standards are written in International English, not American English. This is the English language as it is used in most of the world outside the United States. The most obvious difference is many minor variations in the way words are spelled, for example "colour" rather than "color", "centre" rather than "center" and "behaviour" rather than "behavior." Although ISO granted a special exception to allow VRML to be published initially in American English if necessary to expedite its publication, both the VAG and JTC 1/SC 24 decided to convert most of the document to International English. The only exceptions were changes that would have affected the syntax of a VRML file, such as node names like "Color" and "ColorInterpolator", where a change to "Colour" and ColourInterpolator" would have made existing VRML files incompatible with the new standard.

Use of stylized specification language forms. Those information technologists who work as systems engineers, especially on government contracts, are already familiar with the stylized use of words such as "shall", "shall not", "may", "can" and so forth in writing specifications. For example, "shall" indicates something that is required while "may" indicates something that is permitted but is not required. Moving Worlds Specification required some minor modifications to conform to proper specification language form. One change that was necessary in many places was replacing the word "must" (which is not used in International Standards to avoid any potential confusion with external, statutory requirements) with the word "shall."

Distinction between normative and non-normative text and references. It is not uncommon for a specification to contain material that is for information purposes only. In International Standards, such text is called informative or non-normative to distinguish it from normative text that gives those provisions that are required. Also, many specifications refer to other documents, sometimes to incorporate their text without repeating it and other times just to provide additional information. Some cleanup to the Moving Worlds Specification was required to separate informative (bibliographic) references from normative ones and to clarify which text was normative and which was not.

It took the coeditors and others who assisted them almost two months to make all required changes to the Moving Worlds Specification. On August 4, 1996, the resulting final version of the VRML 2.0 specification was released by the VAG at SIGGRAPH 96 in New Orleans, Louisiana, and the CD ballot was started within ISO. Even at this early stage, recognizable improvements in the quality and clarity of the specification due to the work of experienced ISO and IEC contributors was evident to all those involved in the editing work.

Sidebar:

The life cycle of an International Standard involves several stages. These are:

New Work Item Proposal (NP):

Initial Draft (ID): This is the first version of a document that is created by or introduced into a committee

Working Draft (WD): While a standard is under review and development within a Working Group, it may be circulated for comment as a Working Draft many times. In this form, the draft is often incomplete.

Committee Draft (CD): Once a standard is essentially complete and in close to its final form, it can be circulated as a Committee Draft. This is the point at which the "public" typically first "sees" a standard. According to the practices of many national standards bodies (including ANSI in the US), draft standards receive an initial public review at this stage. There is also formal voting on a standard within ISO and IEC at this stage, with an editing meeting to resolve any comments received on the ballot.

Draft International Standard (DIS): Once a Subcommittee (such as JTC 1/ SC 24) believes that it has finished all the technical work an a standard, it is advanced to the stage of Draft International Standard. Again there is formal voting on the standard, but at the Technical Committee level (JTC 1 in the case of Information Technology standards) rather than the Subcommittee level. Only editorial changes are allowed at this stage, and again there is an editing meeting to resolve any comments received on the ballot. In some countries, there is a second public review at this stage.

International Standard (IS): The final product of the process, ready for publication and sale.

The Draft International Standard Stage

The ISO/IEC CD ballot for VRML closed in December 1996. During the ISO ballot period, VAG members Rikk Carey and Gavin Bell as well as Chris Marrin, an author of the Moving Worlds proposal, monitored the simultaneous review of VRML 2.0 that was taking place on the www-vrml list. They documented any issues raised and resolved on the list which suggested that changes should be made in the specification. The consolidated comments from the list, plus some others from VAG members as individuals, were submitted as comments on the CD ballot from the VAG. This was possible because the VAG had been granted Category C Liaison status with JTC 1/ SC 24.

This CD ballot resulted in over 200 pages of detailed technical and editorial comments. Of the hundred or so comments from the VAG, most were minor technical comments. The overwhelming majority of comments were from SC 24 National Bodies (i.e., countries). These comments were mostly editorial in nature and many concerned places where the specification was imprecise or ambiguous and clarification was desired. All comments were resolved at a joint ISO/IEC and VAG editing meeting December 7-9, 1996.

It is again instructive to look in detail at the nature of the comments that were received at this stage in the evolution of the specification and the resulting changes in the document. The major editorial changes in the DIS version from the CD version were:

The concepts clause was completely rewritten resulting in much greater clarity.

More conformance items were consolidated in the conformance clause.

Many confusing statements throughout the specification were clarified.

Illustrations were added to increase understanding of key concepts.

VRML code was added for all the VRML examples. This code can invoked from within the document itself using a VRML-aware browser.

The wording of the document was further refined to reflect ISO terminology.

There were only a few significant technical changes required as the VRML CD text was turned into the DIS text by the ISO editors and the dozen or so ISO and IEC experts who assisted them. These changes were:

+ and - characters became illegal first characters for DEF names.

In the FontStyle node, the family field was changed from SFString to MFString to support ordered list of choices; similar to URL fields.

In the MovieTexture node, the duration_changed eventOut was changed from SFFloat to SFTime to be consistent with AudioClip.

In the NavigationInfo node, a new value for type field "ANY" was added. The meaning is that the browser can do whatever it wants. This was added because several browsers were doing this anyway (misinterpreting the specification) and thus a mechanism was needed to support the existing policy. This change implies that the other NavigationInfo types are now strictly enforced.

In the Viewpoint node an error was corrected in field of view text (uses minimum angle for field of view, not maximum).

Many further additions and changes were made to the conformance clause and to the minimum support requirements.

A variety of clarifications, corrections, and improvements were made to both the Java Scripting Reference Annex and the JavaScript Scripting Reference Annex.

After all editorial and technical comments from the CD ballot were resolved, a detailed response document was prepared describing each of the changes to be applied to the CD text. This was standard ISO and IEC procedure and differs from the manner in which comments were handled "on the fly" during discussions on the www-vrml list.

The simultaneous and coordinated reviews in both the VRML community (the www-vrml list) and within ISI/IEC JTC 1/ SC 24 each had its benefits. List discussions allowed a rapid paced exchange of ideas, but tended to be focused on mostly "shallow" technical aspects of the specification as uncovered contemporaneously by implementers and users. The SC 24 review took a deeper and broader look at the specification as a whole with an eye towards making it more usable by those beyond the initial implementers. The two types of review were complementary, and each resulted in improvements to the VRML specification. The increased clarity and precision that resulted from each of the two types of review will have lasting benefits to both implementers and users of VRML-based systems.

Due to the volume of comments, it took the editors two months to prepare an "editor’s draft". This editor’s draft was further reviewed by individuals who had attended the editing meeting to insure that all of the responses had been correctly applied. This process took several additional months. Following this, text for the Draft International Standard (DIS) was prepared and posted on the World Wide Web by the editors on April 6, 1997.

The VRML Consortium

By the middle of 1996 it was becoming evident that some key vendors who wanted to make a major commitment to VRML technology were uneasy about entrusting the future of the specification to either an informal and ad-hoc group such as the VAG or solely to an ISO standards committee. Since the VAG members had been hand picked by only a few of the original organizers, there were concerns about how representative they were of the broad VRML community. Both the lack of a formal leader and the lack of well defined and documented processes led to the VAG having a perceived lack of credibility in the minds of the large corporations. These organizations would much rather work with a formal standards body or even another corporation where the outcome would be more predictable. It was suggested many times by other corporations that SGI should form its own VRML ARB (Architecture Review Board, much like the OpenGL ARB) in the hopes of establishing a formal body that they could join and work with.

In response to these concerns, a meeting was held during the SIGGRAPH Conference in August 1996 concerning the possibility of establishing an industry consortium to promote VRML. As a result of this meeting, a Consortium Working Group (CWG) comprised of representatives from industry and academia was chartered with establishing the VRML Consortium. An e-mail discussion list was also established for prospective consortium members. The CWG issued a Request for Proposals for creating the consortium, including determining its organization and bylaws. The CWG received and reviewed three proposals. It recommended a plan presented by Deepak Kamlani of Interprise Ventures and Grayson Schlichting of the Broad Alliance for Multimedia Technology and Applications (BAMTA) for establishing the administrative infrastructure for the consortium. Following this selection, the proposal was discussed on the e-mail discussion group for prospective consortium members where it received positive feedback. At this point, planning progressed to satisfying the legal prerequisites for incorporating the Consortium.

In October 1996, members of the VAG assisted the CWG in refining the proposed VRML Consortium by-laws, based on a first draft provided by Interprise Ventures. This draft was based on the by-laws of two successful high-tech consortia ( the International Multimedia Teleconferencing Consortium (www.imtc.org) and The Vendors' ISDN Association (www.via-isdn.org).) The second draft by-laws were discussed on the prospective members e-mail discussion list in November, 1996 and further modifications were made. After agreeing that consensus had been reached, the CWG and VAG voted on November 15, 1996 to approve the by-laws, to form an interim Board of Directors of 15 members, to appoint interim officers, and to incorporate the VRML Consortium as a legal entity. The VRML Consortium, Inc. was established as a nonprofit corporation with the mission of fostering and evangelizing VRML as the open standard for 3D multimedia and shared virtual worlds on the Internet.

In January, 1997, once the VRML Consortium was established and sufficient number of organizations had joined as voting members, elections for a new Board of Directors were held in accordance with the approved by-laws. Control was passed from the interim to the elected Board of Directors at the VRML 97 Symposium in Monterey, California on February 23, 1997. Initial members of the Consortium included thirty five leading Internet companies: 3Dlabs, Inc., Apple Computer, Inc., Axial Systems, Inc., Black Sun Interactive, Inc., Construct Internet Design Co., dFORM Inc., Division Ltd., First Virtual Holdings, Inc., IBM Corp., Integrated Data Systems, Intel Corp., Intervista Software, Inc., Kinetix, Microsoft Corp., Mitsubishi Electric Research Labs, Inc., Netscape Communications Corp., Oracle Corp., ParaGraph International, S3, Inc., SENSE8 Corp., Silicon Graphics, Inc., Sony Corp., Superscape Inc., Template Graphics Software, Inc. and Visible Decisions Inc. Other members included academic institutions (such as the Naval Postgraduate School, Brown University and the San Diego Supercomputer Center) and governmental entities (such as the USA’s National Institute of Standards and Technology). During the first Board of Directors, the VRML Consortium accepted and endorsed the existing Cooperative Agreement with JTC 1 as negotiated by their predecessor, the VAG.

The Present

Recall that the CD ballot for VRML closed in December 1996 and the coeditors worked for four months, producing the text of DIS 14772-1 in April 1997. This is the current version of the specification that is posted at the VRML Consortium WWW site (http://www.vrml.org/Specifications/VRML97/DIS). The DIS ballot has been initiated within ISO/IEC JTC 1, and will complete in early October 1997. The DIS text has been posted for review by the VRML community since April, 1997, and comments from the www-vrml list are being collected by co-editor Rikk Carey. As with all VRML community comment periods, this one is not as formal as an ISO/IEC ballot period, but is on-going and results from mail on the list which reveals deficiencies in the way in which the VRML functionality is presented. There are many fewer comments now than during the CD review, attesting to an improvement in quality of the specification.

A final editing meeting is scheduled to prepare the text for International Standard VRML. Comments from the www-vrml list will be consolidated and resolved again along with ISO/IEC National Body comments in a similar manner as was done for the CD ballot. It is anticipated that ISO VRML will be published near the end of 1997.

How the Community was Formed

The VRML community is formed by a diverse group of participants. While there is representation from the scholarly computer science research community, there is also wide participation from both the "hacker" community who are used to Internet-style business models as well as from the creative community who design VRML content for the web. There have been minor culture clashes because of this diversity, but the sense of community that has been built by the participants is more important. If we look at how this community came together and was able to accomplish what it has done in so short a time we find several key factors had an important influence:

The Internet-based e-mail discussion list had the most profound effect and was the single most important aspect defining the community.

Many key individuals were personally committed to the success of the work. This was the second most important factor in its success. In fact, several individuals realized that developing VRML would benefit both their own careers and their companies at the same time. The synergy from this alignment of corporate and personal goals was serendipitous.

In the early days of VRML, SGI was the only serious corporate participant. Fortunately SGI management gave the early participants (Rikk Carey and Gavin Bell) a great deal of latitude. As typically happens when a corporation participates in a standards effort, some inside SGI were not so eager to give away technology. One reason that the Inventor technology was submitted to the VRML community and made available for public use was that SGI may have undervalued the technology and thus felt there was minimal risk in giving it away.

SGI must be credited for realizing that anything that broadens the use of 3D (whether on an SGI platform or on a PC) would be good for SGI in the long run. They had had faith that the work would eventually create opportunities for SGI (and others.) An often used phrase in this regard is "a rising tide floats all boats." The same can be said of some other corporate participants who also valued long term market growth over immediate profit.

The computer science research community (notably the ACM’s Special Interest Group on Computer Graphics (SIGGRAPH)) played a minor but key role in providing a meeting place and sponsorship for some key conferences.

Because of the above factors the VRML community was able to coalesce and work together. While a few of the participants had worked together before, most became acquainted during the development effort, so it was not a case of a team that had worked together successfully before being applied to a new problem.

Lessons Learned

The development of the VRML standard is one of very few successful examples of an "Internet" standard being processed to become an ISO/IEC standard. Initially, considerable effort was expended by the leadership of both the VRML community and of JTC 1/SC 24 to overcome the traditional hostility and clash of styles between the Internet community and ISO and IEC. Today, each community has come to realize the value that the other contributes to the standardization process and how a coordinated approach based on combining the best that each community has to offer can be an effective model for standards development. The three most important lessons we have learned are:

Value added by the Internet style

In the early and intensely creative stages of the development of a specification, the use of Internet style discussion lists was very valuable. The VRML Community used this technique to great success in soliciting proposals, evaluating them and refining their initial products. About fifty or so individuals (out of several thousand who were actually on the list) actively contributed at any one time to any one thread of discussion. Individuals from all major interested vendors also participated, and this helped build worldwide consensus. VRML was developed as an open standard instilled with the principle that high quality infrastructures can be built in the open on level playing fields. The result has been faster and better products as well as more interesting and productive market competition based on value, not history or platform dependence.

Value added by ISO and IEC style

The product of an Internet style development is often referred to as "rough consensus and working code." At the point that the experts of JTC 1 got involved, it was evident that this was a fair characterization of the VRML 2.0 Specification. The VRML Community had created an excellent specification based on the right ideas, but one that still needed considerable editorial and technical work to bring it up to the quality expected of an ISO/IEC standard. Since many parts of the VRML 2.0 draft specification were based on VRML 1.0, which was in turn based on one particular implementation (Open Inventor), the experts of JTC 1/ SC 24 worked very hard to increase the descriptiveness of the early working drafts of the VRML 2.0 Specification. The first result was CD quality text produced by the ISO and VRML Community co-editors with about 2 months effort.

The step to DIS required considerably more effort and clearly showed the value of ISO/IEC processes. At the stage that a standard moves to DIS it is very difficult, if not impossible, to accomplish the necessary work by an Internet style discussion list. What is required is long, careful, thoughtful review by independent groups of experts whose comments are consolidated and resolved at an editing meeting (such as the ones held by ISO/IEC). In the case of DIS 14772-1, this took a three day meeting by about a dozen experts to process the roughly 200 pages of comments, followed by four additional months of work by the coeditors to make the agreed changes.

Usefulness of cooperative agreements

The development of VRML was accomplished under a written cooperative agreement negotiated by the parties at the start of their relationship. The use of such an agreement proved key to keeping the work on track and resolving problems at several points in the process. Among the most important elements of the agreement was the constraint that no unnecessary technical changes would be made to the base document. Although at least four significant technical changes were actually made, all were recognized by both parties as being necessary. A second key constraint was that the work was to be accomplished on the shortest feasible schedule. Finally, the agreement makes ISO/IEC and the VRML Consortium "co-owners" of the product. This was very important to the Internet community, whose history and traditions require that their standards be freely available over the Internet.

The Future

In June 1997, JTC 1 and the VRML Consortium mutually adopted the first revision to the original Cooperative Agreement under which VRML was processed as an ISO/IEC standard. Under the revised agreement, JTC 1/SC 24 and the VRML Consortium will continue their cooperation to develop specifications for other areas of VRML technology and to make these into International Standards where appropriate. A key feature of the agreement is that the early technical work will be accomplished principally within the Consortium’s Working Groups (where ISO/IEC experts as well any other individuals are free to participate without necessarily being Consortium members) while the later stage technical and editorial work will be accomplished within JTC 1 (with liaison from the VRML Consortium.)

Early in the development of VRML, it was decided that more ambitious features—such as multi-user interaction or autonomous creatures that can sense and react to their environment—would not be part of the initial version of VRML. VRML Consortium Working Groups (with participation of JTC 1 experts) are currently busy specifying, implementing and testing extensions in many areas. (A list of all approved working groups can be found at: http://www.vrml.org/WorkingGroups/.) The best and most widely accepted extensions will become VRML Consortium Standards and then be processed to become International Standards. Among the areas of active current work are:

A Compressed Binary Format for VRML files, emphasizing the reduction of file size for fast transmission, and the simplification of file structure for fast parsing.

An External Authoring Interface between a VRML world and an external environment.

Humanoid Animation to provide a standard VRML representation for humanoids.

It is the intent of JTC 1 and the VRML Consortium that consortium procedures be followed for designing the functionality of new specifications and for deciding whether any particular specification is appropriate for standardization. The output of a consortium working groups can be either be a report not intended for standardization or a specification for additional functionality which may be submitted for standardization. The VRML Consortium has created the VRML Review Board (VRB) to oversee these activities and to decide on the viability of any particular specification. JTC 1/ SC 24 has a liaison representative who participates in the work of the VRB and serves as the primary technical interface between the two organizations. At a management level, the Chair of SC 24 and the President of the VRML Consortium coordinate directly on all issues of mutual concern.

Conclusion

The processing of VRML within the ISO and IEC standards community has marked many firsts for the development of International Standards, including the completion of a project from start to finish in record time (less than 18 months) and the first International Standard published in HTML. The cooperative work of ISO/IEC JTC 1 and the VRML Consortium in developing DIS 14772-1 provides an example of how an industry specification can be processed into an International Standard through a normal standards development process without either unnecessary delay or needless technical change. It also illustrates the substantial value that the technical experts experienced in the development of ISO and IEC standards can add to a specification. In terms of increased clarity and precision, this value will have lasting benefits to both implementers and users of VRML based systems.

Acknowledgments

Mark Pesce and Tony Parisi were responsible for starting the activities that eventually led to the creation of VRML. RenderMorphics donated a copy of Reality Lab to Tony and Mark, which enabled them to realize their vision and build the first prototype. Using RealityLab they realized for the first time that VRML could be a reality on low-end platforms. Other members of the VAG, who along with author Rikk Carey guided the development work during a very critical period, were Gavin Bell, Brian Blau, Jan Hardenbergh, Jon Marbry, William Martens, Tom Meyer, Mitra, Tony Parisi, and Mark Pesce.. Chris Marrin and Gavin Bell were co-authors with Rikk Carey of the VRML Specification, Version 2.0. Koichi Matsuda (Sony) was a key contributor to the Java annex, Jan Hardenberg (MERL) was a key contributor to the JavaScript annex, and many other individuals from the www-vrml list made contributions of parts of the document. SGI donated the Open Inventor file format and tools to the VRML community and contributed significant engineering efforts to the development of the standard. Finally, the work of the VRML Consortium and its President, Neil Trevett, in making VRML a reality in the marketplace must be recognized.

Within the ISO and IEC community, many people contributed to the success of VRML. Authors Steve Carson and Dick Puk were assisted within SC 24 by Lofton Henderson, the Metafiles Rapporteur and the many experts of the Metafiles Rapporteur Group of JTC 1/SC 24. The Chair of JTC 1, Mary Ann Lawler and the staff of the Information Technology Task Force at ISO Central Secretariat, notably Keith Brannon, fostered the cooperative atmosphere which allowed VRML to be advanced so quickly while still following ISO and JTC 1 procedures.

The Authors

Rikk Carey (rikk@best.com) is President of Wasabi, Inc. of Los Altos, California. Formerly he was Director of Engineering at Silicon Graphics, Inc of Mountain View, California. He is co-editor of ISO/IEC DIS 14772-1, VRML and co-author of "The VRML 2.0 Reference Manual" published by Addison Wesley.

George S. Carson (carson@siggraph.org) is President of GSC Associates Inc. of Las Cruces, New Mexico. He is the Chairman of ISO/IEC JTC 1/SC 24, Computer Graphics and Image Processing.

Richard F. Puk (puk@igraphics.com) is President of Intelligraphics Inc. of Carlsbad, California. He is co-editor of ISO/IEC DIS 14772-1, VRML.

References

Carey, Rikk, The Annotated VRML97 Reference Manual, Addison-Wesley, Reading Massachusetts, 1997, (http://www.best.com/~rikk/Book).

Hartman, Jed and Josie Wernecke, The VRML 2.0 Handbook, Addison-Wesley, Reading Massachusetts, 1996.

International Organization for Standardization and International Electrotechnical Commission, Procedures for the Technical Work of ISO/IEC JTC 1 on Information Technology, Geneva Switzerland, (Third Edition, 1995) (http://www.iso.ch/dire/jtc1/directives.html).

International Organization for Standardization and International Electrotechnical Commission, ISO/IEC Directives, Part 3, Rules for the structure and drafting of International Standards, (Third edition, 1997) (ftp://ftp.iec.ch/pub/.contents.html).

ISO/IEC JTC 1/SC 24/N1666, "Comments and Responses on CD ISO/IEC 14772-1", March 1997.

ISO/IEC JTC 1/SC 24/N1575, "CGM Rapporteur Group Meeting, Kyoto Japan, 14-18 June 1996", June 1996.

VRML Architecture Group, the International Organization for Standardization and the International Electrotechnical Commission, The Virtual Reality Modeling Language Specification, Version 2.0, ISO/IEC CD 14772-1, 4 August 1996.

VRML Consortium, the International Organization for Standardization and the International Electrotechnical Commission, The Virtual Reality Modeling Language, ISO/IEC DIS 14772-1, 4 April 1997.

Wernecke, Josie, The Inventor Mentor: Programming Object-Oriented 3D Graphics with Open Inventor, Release 2, Addison-Wesley, Reading Massachusetts, 1994.

Previous Features