Open Architecture Theme of TRON Symposium

--TRON Association Celebrates its 10th Anniversary--

Open Architecture and TRON

Open computer architecture and the TRON Project's role in promoting it and cooperating with other open architecture movements was the theme of the 14th TRON Project International Symposium, which was held in Sanjo Hall on the campus of the University of Tokyo on Thursday, March 12, 1998. As if to emphasize the staying power of this powerful idea of openness in computer design, on the following day, Friday, March 13, 1998, the TRON Association held a commemorative ceremony to celebrate the 10th anniversary of its founding.

Following an opening address by the TRON Association General Chairman Koji Nihei, TRON Project Leader Prof. Ken Sakamura kicked off the 14th TRON Project International Symposium by delivering a keynote speech on open architecture and TRON. Since the TRON Architecture is aimed at the computerization of human society, open architecture has been an important concept in the TRON movement since its inception, Sakamura pointed out. In the TRON Architecture, the open architecture approach is based on three main concepts; these are: (1) loose (or weak) standardization in which only the portions that have to be standardized for the sake of compatibility are specified; (2) division of the architecture into levels so that it can be scaled up or down to meet the needs of large numbers of applications; and (3) programmable interfaces so that networks can be upgraded dynamically. Prof. Sakamura pointed out that up to now the TRON researchers have spent a lot of time on the standardization of of ITRON and CTRON at the lower levels. However, in the future, development efforts will shift in the direction of standardization of the human-machine interface and rules for data exchange, he said.

The TRON Project is not the only computer movement in the world in which the idea of "openness" reigns supreme. The Internet millions of people so eagerly use today is based on an open specification, and there are free operating system movements such Linux and FreeBSD, which open the source code of the operating system and applications to users. A member of the FreeBSD movement, Mr.Satoshi Asami of the University of California at Berkeley, attended the 14th TRON Project International Symposium to participate in a panel discussion titled "Obstacles for Open Architecture" with Prof. Noboru Koshizuka of the University of Tokyo and Mr. Akira Matsui of the Personal Media Corporation, both of whom work on the BTRON architecture for personal computers and workstations.

Panel Discussion on Obstacles for Open Architecture

The panel discussion began with the chair, Prof. Koshizuka, expounding on the meaning of "open architecture." Prof. Koshizuka said open architecture is good for three reasons: (1) it follows the tradition of keeping all intellectual information open, (2) it stimulates technological development even in closed architectures (i.e., today's standards), and (3) it allows for fair competition in which no vendor has secret inside information. Moreover, he pointed out that since an open architecture is usually created by a large, open community, which in many cases is international in nature, an open architecture will obtain a large number of reviewers, developers, contributors, and users all over the world. At present, for open architecture people there are high-level software libraries (for graphics, networking, etc.), high-level software development environments, and hardware with high-level interfaces. Unfortunately, most open architecture systems run on IBM-PC/AT compatibles, which is where they run into trouble. This is because there are many closed hardware specifications that have become de facto standards on the IBM-PC/AT architecture, which is not to mention the fact that there are minor variations in the workings of even basic parts of the IBM-PC/AT compatibles. The second speaker in the panel discussion, Mr. Satoshi Asami who is the Ports Manager for the FreeBSD movement, went into this issue in further detail.

FreeBSD is an operating system for IBM-PC/AT compatibles (including NEC Corp.'s PC98 series) based on the Berkeley Software Distribution version of AT&T Corp.'s UNIX operating system, the current version of which is based on 4.4-Lite2. Since even the source code of FreeBSD is given away free of charge without any strings attached--making it possible to modify it and sell it--the project is run entirely by volunteers, although some of these make a living from the their FreeBSD expertise by doing consulting, creating CD-ROMs, marketing pre-installed systems (turnkey systems), and offering Internet services as Internet service providers (ISPs).

A major problem that the FreeBSD movement has encountered in recent years, Mr. Asami said, is that most computer parts manufacturers only provide device drivers for major commercial operating systems. That leaves free operating system proponents in the position of having to write their own drivers, which usually end up becoming obsolete after a few months when a new part hits the market. Moreover, Mr. Asami pointed out that some parts manufacturers don't even provide enough information for someone write their own driver--Japanese companies being worst in this respect, he added. So what can be done to overcome this obstacle to open architecture? Mr. Asami said the following steps are available for those who need a device driver or the information for writing a device driver on their own.

Keeping asking the manufacturers nicely (and particularly in large groups)
Sign a non-disclosure agreement (NDA) and write the driver for yourself only
Reverse engineer
Pay a consultant such as BSD Inc. (BSDI) to write it for you
Boycott uncooperative parts manufacturers

As a result of the above measures, some computer part manufacturers--makers of SCSI adapters and video cards--changed their policies and instantly became favorites in the BSD community, he said. Others, he noted, keep changing their interfaces, which has led to their products not being recommended to users. Mr. Asami said the biggest problem for FreeBSD at present is the lack of commercial software. That's not to say that there is no software available for the operating system. He pointed out that FreeBSD CD-ROMs include source code and pre-compiled binary code for over 1,000 third-party software applications. Moreover, there are also emulators to run software written for the highly popular Linux operating system, and also for Microsoft Corp.'s Windows 3.1 and Windows 95 operating systems. However, the latter two emulators do not work very well because companies like Microsoft never go out of their way to share data.

The third panel speaker, Mr. Akira Matsui of Personal Media Corp., spoke in depth from the Japanese perspective about the difficulties of porting an alternative operating system--in his case, the BTRON-specification 1B series--to IBM-PC/AT compatibles (referred to as DOS/V machines in Japan). Some of the problems his company has encountered are similar to those experienced by Mr. Asami and his FreeBSD associates. These result from the fact that the IBM-PC/AT compatible architecture originally had no high-resolution graphics modes, no sound function, no small computer system interface (SCSI) bus interface, no networking functions, etc. In many cases, third-party proprietary products were developed to fill those gaps, and these later developed into de facto standards.

Other problems, Mr. Matsui pointed out, are unique to the Japanese personal computer market; they result from the fact that in the 1980s major Japanese electronics companies developed multiple incompatible hardware architectures for running the same operating system, namely Microsoft Corp.'s MS-DOS. Thus there exist specialized parts from these previous architectures in today's DOS/V machines. In addition, he has discovered slight differences in timing and hardware bugs; these did not bother the non-multitasking Windows 3.1, but for the multitasking 1B series they proved to be a headache. Some of the fine differences in DOS/V specification personal computers that caused difficulties when porting the 1B series are as follows:

IBM (ThinkPad): the detection bit used when a FD is inserted has the reverse meaning
Dell: method of switching the setting of 2HD/2DD FDs is different from other PCs
Olivetti (Quaderno): can only boot from the first partition when multiple partitions are used
NEC Aile (Note): power save function changes color palette on its own when cover is closed
Parallel port: difficult to detect unconnected printer since bit that indicates connection varies
1.2MB 3-mode FDD (Japan): support method varies with maker; specification unclear

In addition to these fine differences, Mr. Matsui said there are two bad trends in the DOS/V architecture. These are: (1) hardware is becoming more complicated and intelligent, which makes it difficult to understand specifications and write device drivers; and (2) the appearance of Macintosh-like Plug and Play features, which have to be set from the MS-Windows partition due to the complicated nature of the operations of the boards on which they are based. On the other hand, he pointed out that there are some good trends also in the DOS/V world. Mr. Matsui noted that the power saving function for DOS/V machines has been standardized with APM and ACPI. He said that open standards for connecting peripherals have also appeared in the form of Universal Serial Bus (USB) and Peripheral Component Interconnect (PCI) serial bus. In addition, some companies, such as Cirrus Logic, are making hardware specifications public via the Internet; and finally, there are many other non-Microsoft operating systems that are competing with Microsoft Corp.'s Windows on DOS/V machines, he said.

After Mr. Matsui spoke, the three panelists discussed how to get the specifications for various pieces of hardware and what the future holds for open architectures. The three ended up agreeing that getting the proponents of the various open architectures to work hard on the writing of device drivers and sharing that knowhow with other open architectures is the key to their future survival.

BTRON Technical Session

Following the panel discussion on obstacles for open architecture, there was a technical session on the BTRON subarchitecture during which two speakers, Mr. Akira Matsui of Personal Media Corp. and Mr. Shinsuke Yura of the Sakamura Laboratory of the University of Tokyo, respectively gave presentations on the multilingual function and the Multimedia Multi-User Dungeon (MMUD) for BTRON-specification computers.

One of the features that distinguishes the BTRON-specification computer from other personal computers is that it was designed from the beginning to process multiple languages. This is why it is attracting considerable attention as a platform for browsing the World Wide Web, and also as an environment for computer-assisted foreign language learning. Mr. Matsui first described the current implementation of the BTRON multilingual system, which he referred to as a "casual implementation," and then contrasted it with the full implementation, which he called the "thorough implementation." The casual implementation uses a single, fixed plane that can handle a maximum of 48,400 characters. In practice, it handles the standard character sets used in China, Japan, and Korea (a total of 22,024 characters) plus the Japan Industrial Standard (JIS) auxiliary set (6,067 characters) and six- and eight-dot Braille (64 and 256 characters, respectively); moreover, it is capable of mixing multiple language strings.

The thorough implementation, which is scheduled to be implemented on a forthcoming version of the BTRON-specification operating system that has been given the tentative name "3B/V," will use a language specifier code (0xFEnn) to switch "language planes" and thus language processing algorithms. The thorough implementation can handle a virtually unlimited number of characters, but the largest character set it will actually handle is a kanji character set currently under development that lists a maximum of 100,000 characters. Although this may seem like too many characters to a westerner, this large number of characters is needed in Japan for processing place names and personal names, which is not to mention creating digital archives at Japanese libraries and museums. Since the character set is so large, a kanji database is being created in parallel to manipulate it properly.

TRON Project Leader Ken Sakamura's main job now is running the Digital Museum at the University Museum of the University of Tokyo, which has been an outstanding success. His Web site at the University of Tokyo gets an average of 500,000 hits per month, the majority of which are to access the digitized materials at the University Museum. In order to allow users to walk through the three-dimensional virtual space of the rooms in the Digital Museum, a Multimedia Multi-User Dungeon (MMUD) function has already been created and demonstrated. Now, to enhance this virtual experience of walking around "exhibit objects" (digitized images of physical exhibits), a method of inserting "player objects" (images of the bodies of the people visiting the virtual museum spaces) has been developed. The method is referred to as "augmented virtual reality," and it was described by Mr. Shunsuke Yura, a graduate student studying under Prof. Sakamura.

The new augmented virtual reality system uses the "chromakey technique" to clip and embed into the three-dimensional virtual reality space an image of a player's complete body, thus allowing for the player to use normal gestures in communicating with other players. The system takes video images in from either a real-time video source or a recorded video source and feeds them into a video server connected to MMUD browsers. The video server is controlled via either a real-time control device or a video control script. The video control script does the following: (1) it moves the embedded image, (2) it navigates the player through virtual space, (3) it provides a paint function (that also supports text) for drawing/writing comments, (4) it embeds hyperlinks (either a "warp link" that brings the player in front of the virtual object the link points to, or a "tool link" that allows the player to examine a exhibit object in detail), and (5) it merges multiple video images.

CTRON Technical Session

The CTRON session at this year's symposium was unique in that the only speakers to give presentations on the CTRON subarchitecture were two speakers from Shanghai, China: Mr. Sun Ting Cai of China Shanghai CONTEC Co. and Prof. Shen Jianhua of East China Normal University. Shanghai is the location of the world's first overseas TRON association, which was established after the TRON Association held three seminars there to introduce the ITRON and CTRON subarchitectures starting about four years ago.

Mr. Sun Ting Cai spoke on behalf of the Shanghai TRON Association and said that the TRON seminars were very welcome in Shanghai, and many products based on ITRON and CTRON have been announced. ITRON has been put to use in an elevator control system and CTRON is being used for a data acquisition system in factories. However, TRON is still not widely known in China. Mr. Sun said that when he went to the Japanese consulate to obtain a visa and wrote "TRON" on the application form, the consulate official thought he was going to Japan to do something with "clones." Mr. Sun said, "no problem, I'm going to clone TRON and bring it back to China."

After Mr. Sun spoke, Prof. Shen spoke about a CTRON-based real-time data acquisition system he developed for use in collecting computer-controlled assembly line data (e.g., temperature, pressure, etc.) in factories. Prof. Shen's system is a distributed system based on a master/slave architecture in which an Intel 486-based master computer is connected via an RS-485 serial interface to 12 slaves. The slaves are all "intelligent instruments" equipped with either a microprocessor or a microcontroller. The master collects regular measurements from the slaves in the distributed system system, which has a response time of less than 100 milliseconds, and puts them into a database. Using a keyboard command, statistics can be tallied and analyzed, and they can be output with a print command.

Mr Sun then took the floor and described the universal intelligent controller (UIC) for industrial processes that his firm has developed using the CTRON operating system. The universal intelligent controller is a two-part hardware module about the length of a ball point pen, which is based on a 16MHz Intel 80C286 microprocessor and a 16MHz 80287 coprocessor on the motherboard. The CTRON operating system is burned into firmware, and the module has an ISA bus interface so that it can fit into any standard IBM-PC/AT compatible based on an Intel 486/Pentium equipped with either Microsoft Corp.'s Windows 3.1 or Windows 95. Although MS-Windows is used as the human/machine interface, all other processing is done by the CTRON-based module, thus making a standard PC a multitasking system for industrial use. Mr. Sun said that his universal intelligent controller has been applied in power plants, steel mills, and other industrial fields in China.

Mr. Toshikazu Ohkubo of NTT Software Corporation, who chaired the session and also played a leading role in holding the TRON seminars in China, pointed out that CTRON is used mainly in telecommunications and financial service applications in Japan, but in China it is being put to use in industrial production and factory automation.

ITRON and JTRON Technical Session

The ITRON session began with Prof. Hiroaki Takada of Toyohashi University of Technology describing the second phase of the ITRON subproject along with the fourth version of the ITRON specification, i.e., micro-ITRON4.0. As a result of the rapid spread of ITRON-specification kernels throughout various Japanese industries, in its second phase the ITRON subproject must specify "related standards" for embedded systems, in particular software components, development environments, and application-specific standards. The next generation micro-ITRON-specification kernel is being designed in accordance with these standards; it will have improved software portability and incorporate new kernel functions for hard real-time support for software components with hard real-time constraints, such as software modems and MPEG-based video functions.

Software components have increased in importance for embedded systems as they have grown in size and complexity, Prof. Takada said. This is because development from scratch becomes more and more difficult as time goes by, and a lack of expertise can be a serious problem in trying to program functions one is not familiar with. A digital camera is a good example of this. Although the company that makes the camera may know how create applications to take and store digital pictures in JPEG format, it might have difficulty create a function to send them through the Internet, which requires in-depth knowledge of the Internet's TCP/IP protocols. However, if that transmission function is available as an off-the-shelf function component, development time and costs are greatly reduced. Accordingly, ITRON subproject researchers plan to standardize APIs for software components in specific fields, and then develop and circulate software components based on those standardized APIs.

The micro-ITRON4.0 specification is being developed to advantage of higher performance processors; it is also introducing the concept of "profiles," which are collections of functions for various applications. Like its predecessor micro-ITRON3.0, the micro-ITRON4.0 specification will be divided into three levels. In comparison to micro-ITRON3.0, which is divided into required, standard, and extended levels, micro-ITRON4.0 will be divided into minimum, vehicle control, and standard levels. The subset for vehicle control was proposed by automotive engineers who need a subset that offers a smaller and higher performance profile than the current standard subset of micro-ITRON3.0. In addition to the kernel specification itself, application design guidelines for real-time systems will also be drawn up. There are two reasons for this: (1) to guarantee real-time constraints of both the software components and applications (rate monotonic analysis [RMA] is adopted for this purpose), and (2) to provide people who are new to the field of real-time system design with a good set of design guidelines.

Prof. Takada said that new study groups and committees have been set up to deal with the above-mentioned standardization activities. These are: the ITRON Hard Real-Time Support Study Group, the Embedded TCP/IP Technical Committee, the RTOS Automotive Application Technical Committee, and the Java Technology on ITRON-Specification OS Technical Committee. In 1998, these groups will publish the following documents: the micro-ITRON4.0 Specification, ITRON TCP/IP API Specification, Application Design Guidelines, and JTRON ver.2 Specification.

On the topic of JTRON, Mr. Yukikazu Nakamoto of NEC Corp. spoke on the subject of integrating Java and the ITRON kernel. The Java language has had quite an impact in the world of computing because it is a simple, object-oriented language with strong type checking that can be used to write programs that run on any computer. However, from the point of view of the TRON Project, which aims at the creation of real-time networks, it presents a big problem in that it is not suitable for real-time applications. Java's slow execution speed, its garbage collection function which prevents predictable action, and its inability to access specific low-level hardware functions all work against in real-time environments. To overcome these problems, the ITRON Technical Committee is studying a hybrid approach called JTRON in which the best parts of Java and ITRON are merged together via a cooperative framework. Three methods have been studied as a means of creating a standardized framework: (1) a socket approach in which the Java stream communicates with ITRON tasks, (2) a conversion approach in which Java applets are are handled as ITRON tasks (Java threads become common objects, which are given ITRON IDs and processed under lock control), and (3) wrapper processing of ITRON tasks as a Java attachment class. The JTRON specification is scheduled to be made public in May 1998, Mr. Nakamoto said.

Panel Session on Potential of Partnership with Other Platforms

The TRON Project has been maligned many times by uninformed--and sometimes utterly disingenuous--sources claiming it is a "trade barrier," or that it "ignores current standards." However, the fact of the matter is that TRON is an "open architecture," and as such it can facilitate trade, since anyone is free to create TRON-compatible products without the payment of royalties. The same can not be said for today's "market leading computer architectures." Moreover, the TRON Project has from the very start made provisions for current standards, particularly when they are open standards. That is why, for example, the CTRON specification defines interfaces for four of the seven layers of Open System Interconnection (OSI). Against this background then, it should come as no surprise that the second panel session at the 14th TRON Project International Symposium was on the topic of "TRON's Potentiality of Partnership with Other Computer Platforms."

The first speaker was Mr. Yukikazu Nakamoto of NEC Corp., who spoke about his work of merging ITRON with the Java programming language and CORBA. Mr. Nakamoto is very experienced in merging ITRON with non-TRON technologies; he previous created a real-time version of UNIX by merging UNIX with ITRON. An example of this real-time version of UNIX is scheduled to be placed in orbit in 2001 as part of the international space station project. CORBA, for those who haven't heard of it, is part of a new object-oriented technology called the Open Management Architecture (OMA), the goal of which is to enable different makes of computers to operate together in networks. CORBA, which stands for the Common Object Request Broker Architecture, is the heart of OMA; it is the system that allows the network objects to converse with each other, regardless of which platform they originate from or what technologies were used to implement them. OMA is being developed by the Object Management Group, a non-profit organization established in 1989 that currently has over 700 members. Mr. Nakamoto said the reason for merging TRON with these other platform is that it becomes easier to create applications; moreover, it also enables existing systems to operate cooperatively in TRON-based systems.

After Mr. Nakamoto spoke, Prof. Noboru Koshizuka of the University of Tokyo spoke on the subject of merging the hypermedia filing system standard on BTRON machines with the hypermedia filing system of the Internet, i.e., the World Wide Web. Although at first glance the BTRON filing system and the WWW seem to be one and the same, Mr. Koshizuka said the two are in fact different. He summarized the differences between the BTRON filing system, the real object/virtual object model (RO/VO), and the WWW as follows: (1) the RO/VO model links in both directions, while the WWW links in one; (2) the linking point in the RO/VO model is a region, but a point in the WWW; (3) a window to another page is an opened virtual object in the RO/VO model, but a frame in the WWW; and (4) transition to a new page is a new window in the RO/VO model, but it takes place in either the same window or a new window in the WWW. In order to combine BTRON's filing system with that of the WWW, Mr. Koshizuka said the following is necessary: (1) a method must be devised to package the multirecord files used in BTRON (the data format of TAD [TRON Application Data-bus]) into a single file; (2) the URL (uniform resource locator) of the WWW must be represented in the virtual object (the link to a BTRON file) data; and (3) the real object (the BTRON file itself) must be represented in the URL.

Mr. Ryu Koriyama, chief executive officer of Aplix Corporation, then gave a description of his company's commercial real-time operating system based on the JTRON specification, which is called "JBlend." Mr. Koriyama previously worked as the project manager for the Japanese language-version of Microsoft Corp.'s OS/2 operating system, which puts him in a slightly awkward position inside the TRON movement. However, he is now doing good things for the TRON Project with a team of about 70 programmers, who have already implemented three versions of JBlend for ITRON1, micro-ITRON2, and micro-ITRON3. JBlend is implemented using a Processor Abstraction Layer (PAL) through which all the CPU-dependent portions of the Java operating system routed into the ITRON real-time multitasking kernel. Moreover, all of the Java application program interfaces (APIs) are mapped to the ITRON APIs. As a result, all of the functions of ITRON are realized in JBlend. JBlend-based products have already been developed and exhibited in the U.S. These include "Blue Mountain," a handheld remote control terminal developed by PFU; an Internet-compatible color facsimile machine developed by Panasonic; and JBlendPC for IBM-PC/AT compatible personal computers.

Unfortunately, there are some areas where it is difficult for the TRON Project to cooperate with other operating systems. One of these is in the area of character codes. The BTRON subarchitecture is soon scheduled to be blessed with the richest character set ever developed for a computer system. Since it will have more characters available, and even be able to distinguish among languages text is written in, exchanging data with non-BTRON-specification computers will pose some problems, particularly at the non-TRON receiving end. Accordingly, when asked about this problem, Prof. Ken Sakamura said, "In order to solve this [problem], when bringing in data from the outside world, it is necessary to either save information that you cannot interpret, or to make it so that it can be expanded with a data description language or something [along those lines]. In TRON, fusen and TULS can realize this; it is necessary for other open types [of computer platforms] also to prepare a similar mechanism for us."

Another trend in cooperating with other platforms is bringing other platforms' technology to the TRON Architecture. Prof. Hiroaki Takada, the panel discussion moderator, described how the Internet's TCP/IP communication protocols are being implemented for the ITRON specification. In the world of UNIX, where they originate, they are handled as a socket library. However, it has been determined that bringing them over as is from UNIX is inefficient, so they are being converted into a separate TCP/IP stack. In addition, Prof. Takada talked about how the ITRON subproject has agreed to cooperate with a TRON-style automotive real-time operating system (RTOS) project in Europe called OSEK/VDX. Unlike the ITRON subproject, which aims at creating a RTOS that can be used in a wide variety of applications, the OSEK/VDX project is more focused. It is aimed at the creation of an operating system specification for networking the various control systems inside an automobile, plus standardized communication and network management protocol APIs. Like the TRON Project, OSEK/VDX began as a joint project between industry and academia in 1993. Prof. Takada said ITRON researchers learned about it around 1996. Even though it is in a way a rival project, since their goals were similar and their project is open, a cooperative effort was launched. There are two developments from the OSEK/VDX project that will help the ITRON subproject, Prof. Takada said. These are: (1) communications and network management specialized for automobiles, and (2) a standardized interface for use between debugging tools and a RTOS. In the next generation, both projects plan to cooperate more closely.

After Prof. Takada spoke, Mr. Toshikazu Ohkubo, chairman of the CTRON Technical Committee, gave an unscheduled rundown on the Multimedia Network Service Platform (MNP). The development of this system for providing centralized multimedia services in networks was commissioned by the Information-technology Promotion Agency, Japan (IPA) as part of its Advanced Software Enrichment project. MNP is being developed on top of a CTRON-specification operating system, Fujitsu's ROSEC kernel, and it makes use of both CORBA and Java for providing services. He said this system is now being evaluated to see how well a distributed platform (CORBA) runs on top on a real-time operating system. Up to now, CORBA-based distributed systems have mainly been run on UNIX-based systems.

The TRON Association Celebrates its 10th Anniversary

The TRON Association celebrated the 10th anniversary of its founding with a series of commemorative lectures and a reception on Friday, March 13, 1998, the day after the TRON international symposium. The lectures covered "Computer Augmented Environments" and the "Multimedia Network Service Platform"--two projects based on TRON technologies commissioned by the Information-technology Promotion Agency, Japan--plus "Communications in the 21st Century," a guest lecture gvien by Prof. Atsushi Matsushita of Keio University. In addition, TRON Project Leader Ken Sakamura delivered a special lecture on "10 Years of the TRON Project," and Prof. Hiroaki Takada moderated a panel discussion on the "Standardization and Future Prospects of Embedded Software." The panelists in the panel discussion session included specialists from Toyota, Mentor Graphics, Mitsubishi Electric, and Sun Microsystems.