At its meeting on Tuesday 4 April the proposed strategy and detailed workplan (1989/90 and 1990/91) for EASE was approved by the Computing Facilities Committee (CFC). The Workplan includes the effort available at the Rutherford Appleton Laboratory and additional effort to be made available through development contracts.
The strategy for EASE takes due account of the Engineering Board's priority areas identified in the new SERC Corporate Plan, namely:
The EASE mission statement for the community to associate with is:
To stimulate and encourage engineers to use appropriate state-ofthe-art software and hardware to enhance their research.
The three major planks for the programme of work are:
A special edition of the Newsletter will be published next month, giving full details of the EASE strategy and workplan.
Application Software and Technical Report for Academia is a joint project between IBM and CNR (the Italian National Research Council). The aim is to design and implement a free European service through EARN (European Academic and Research Network) for the distribution of information, software in public domain and documentation about academic projects in education and research.
The project started in January 1988 and has a 2-year duration. Initial data will include
The service allows EARN users to retrieve documents located in the ASTRA database at the CNUCE (Centro Nazionale Universitario de Calcolo Elettronico) Institute, Pisa, or in other databases located elsewhere but accessible to ASTRA.
STAIRS/CMS has been chosen for the creation of the database. The interface is written in REXX language, making the service accessible by EARN users with standard VM/CMS only. In the future it is planned to implement other interfaces in order to render the service available to all EARN users.
The UK access to the service is planned for 2nd quarter 1989.
The major advance in user interface design in the computing area has been the availability of window-managed systems. With window-based systems, the display area is separated into areas known as windows which are logically separate, and rather than use the keyboard to input commands to the computing system, the mouse point device is used to select items for subsequent operations. The Apple Macintosh is probably the best-known exemplar of such systems.
Underpinning this style is the window management system, which organises the system as a set of resources, including matters such as the manipulation of the windows on the display, the routing of input from devices to the required process(es), and the provision of graphics, text primitives by which information can be written to the window by processes.
There have been many window systems since the early developments at Xerox PARC in the 70s. Recently a large number of manufacturers have collaborated on the definition and production of a (de facto) standard window system called the X Window System, or just plain X. This system has been developed at the USA's prestigious MIT Laboratories, with the intention of providing a system which would not only support distributed applications (ie applications which execute on a number of computers simultaneously), but also which would support such applications on a range of different machines, from Crays to PCs, and with a good user interface. X itself does this by using a Client-Server model of operation: the clients are the application programs, which make requests to the resource controller, known as the server. The resource allocated is the display and input from the various input devices, which must be distributed to appropriate clients. These clients need not reside on the same machine as the server. As well as providing these basic facilities, the X system supports a number of (independently written, for the most part) interaction toolkits which aid the application programmer to provide a user interface. There is a defined, machine-independent protocol by which clients and servers communicate.
The X seminar, held at RAL under the CFC's EASE Awareness programme, was given by the X training specialists, IXI Ltd. The seminar was very popular - over 140 attended, from a range of organisations. Topics covered included: a general overview of the X system and it~ history; the X architectural model; programming with Xlib (the very basic X library); and the use of the Xtk toolkit (which makes direct use of Xlib, by and large, unnecessary). A session on developing software using X and the toolkit finished a worthwhile day. Our thanks to Ray Anderson and Clive Feather for their expert presentations.
The Environment Committee has recently approved a new initiative to encourage and support research into the Application of Information Technology to the Construction and Transport Industries. This initiative, supported by significant additional funding, is a co-ordinated programme of research and development seeking economic returns for British industry through the use of the new informing technologies.
The programme is being co-ordinated for the SERC by Professor James Powell and Dr Roger Day, both from the Centre for IT Applications (CITA), in the Faculty of Environmental Studies, at Portsmouth Polytechnic. They are presently seeking ways of identifying and stimulating proposals which will contribute to a coherent programme of applied research of major value to all aspects of construction and transportation. In this way it is hoped to stimulate direct interest and early support from industry, thereby promoting the dissemination and future uptake of the research. The programme's cross-disciplinary strategy will attempt to exploit IT creatively in the built environment by developing the common base of methodology and technology operating across the whole range of the Environment Committee's interests in building, civil engineering and transportation.
Some of the concepts currently known to be relevant to this initiative include Software Engineering, Intelligent Knowledge Based Systems (IKBS), Computer Aided Design (CAD), Computer Aided Manufacture (CAM), Artificial Intelligence (AI), Information Acquisition, Information Transmission/Interpretation, Information Processing/Storage, Man-Machine Interfacing (MMI) and Information Systems Integration. This list is not meant to be exhaustive but simply gives an indication of potential research areas.
There was a successful meeting at the Royal Institute of British Architects in London on 14 March, attended by 45 academics and 15 industrialists from all walks of environment activities. Attendees worked in three subgroups identifying possible themes and are now developing projects within the programme to match these themes.
The Network Services Seminar was held in R22 Lecture Theatre at RAL on 23 February 1989.
The morning session covered the EASE Server Programme and NFS (by Bob Day, Informatics Department, RAL), which was similar to the talk given at the recent User Meeting (January), and Mail Services to other Networks (by Ian Smith, Network Executive). Ian gave details of the addressing problems which might be encountered in sending mail to foreign networks, especially dealing with the problems of address ordering and naming of gateways. (It is often the case that a reply address cannot be constructed from the From field of the message.) His talk included a list of mail addresses of places where help could be obtained. Copies of this slide were made available to participants, and there was a general view that up-to-date information on this topic should be readily available to the Community.
The afternoon session started with two talks on the use of Remote Procedure Calls from Sun (Larry Kluger) and Apollo (Glenn Brennan). Sun provided a tape of RPCGEN, which can be made available on request to the Community. This software assists in the construction of procedures which are to be called remotely. Apollo described NCS, which is available on a number of different architectures (including Suns), and gave an example of the steps needed to code remote procedures (using their NIDL compiler). A more detailed example was provided in handouts.
The final talk described the Supercomputer facilities in the UK, and the current methods of access (given by Trevor Daniels, Central Computing Department, RAL). He also described future plans, including a proposed fast link between RAL and Daresbury. Interest was expressed in hearing about the results of this work. There was also interest in the results of the Working Party on the Future Use of Cray, due to report in June.
Although attendance was smaller than hoped, there were a number of expressions of thanks. There was also a request for a Seminar on Computer Security at some future date.
Regular readers of the Newsletter will be aware of the Central Server policy of the Computing Facilities Committee (CFC). During the current financial year 1989/90 at least 5 servers will be available for multi-department use. This invitation to submit notice of interest in bidding for one of these servers is the first stage of the placement process.
Your bid should not exceed two sides of A4 and should include brief details of the Faculty Departments bidding, existing and planned SERC (Engineering Board) grants, existing single user systems provided on grants, and from other sources, any servers already operating, local networking facilities and some estimate of the size of the community to be supported.
These preliminary bids will be considered by the Computing Facilities Technical Advisory Group at its meeting on Monday 12 June, when a shortlist will be selected to be invited to submit a more detailed bid on which the final placement decisions will be made.
Your preliminary bid should reach Geoff Lambert at RAL by 22 May and you will be advised of the outcome of this by 19 June 1989. The detailed bids will be subject to a deadline by 31 July.
An Advisory Group in CFD, chaired by Professor Hutchinson of Cranfield Institute of Technology, reported to the Electro Mechanical Engineering Committee (EMEC) in mid-1988. A workshop was held in Abingdon on 4/5 January 1989 to present the report, and to discuss the future programme.
Attendance was by invitation and the number restricted to about 40, with a good mix of academics and industrialists. Most of the talks were on the use of and future development in CFD by leading industrial practitioners. There were two sets of Working Group sessions which made suggestions for priority application areas in CFD.
The workshop recommended that a Coordinator should be appointed to pull together the existing and future efforts in CFD, and should have responsibility for computing support for the CFD community. It agreed with the proposal that CFD Development Units should be identified, covering different areas and application. A number of recommendations on training, software production, and availability of software and data were also made.
Workstations have been widely used to solve 20 and to some extent 3D engineering problems. To handle the additional size and complexity of most 3D applications and the more demanding 2D ones, two problems need to be solved:
These problems make it necessary to investigate the underlying hardware and standards, which are discussed in this article.
In 1988, new workstations came onto the market that considerably speeded up the graphics and computational power available to a user, compared with other workstations. They bring together in one product a high performance workstation, a super computer and a high performance graphics engine. Stellar and Ardent, both start-up companies, have produced workstations of this kind.
Superworkstations allow a new class of problem to be tackled in the interactive design process, that was previously beyond the conventional workstation. For large 20 calculations, most 3D calculations and effective 3D presentation, superworkstations allow these demanding procedures to take place in one package.
To gain some indication of the speeds, the Stellar's theoretical peak performance is 40 MFlops, 20-25 Mips and 150K Gouraud shaded polygons per second. This speed is obtained by combined vector and/or parallel processing architecture with custom-designed graphics rendering processors. What justifies the claim to be the basis of an integrated workstation is the internal data path speed which in the case of the Stellar is 320 MBytes/sec.
Both superworkstations offer widely available software: Unix System V, Fortran and C. The language compilers automatically recognise certain statement sequences that can take advantage of the vector and parallel processing facilities. Some user guidance by means of directives is required for maximum performance.
Currently the graphics software, although powerful, is at the basic library level and would need some study to use it effectively. PHIGS+, about which more is said below (under Standardisation), is offered by Stellar and later this year will also be available from Ardent. Ardent have also developed their own system called Dore, which is available on their Titan. Under a licensing arrangement, it will in the future be offered on other systems, including Sun.
A Stellar GS 1000 and an Ardent Titan will be used at RAL for evaluation and development under the EASE programme. The task is to assess their suitability for engineering research and make that expertise available to the community. To go deeper than issuing reports on their basic capabilities, it will be necessary to gain experience of their use on engineering problems. To some extent this can be done (and is being done) by suitable prototypes at RAL, but exposure to engineering problems from HEIs would make the evaluation deeper and more useful. An individual researcher, who wishes to try out a particular application on this class of workstation, is welcome to use these facilities as a test bed at RAL. This could precede a decision as to which workstation hardware and software to apply for on a more long term basis. Such users will be expected to report on their experiences of the suitability of the visualisation software and hardware for their engineering application, so that others in the EASE programme may gain the benefit of this experience.
Anyone interested in following this up should contact Julian Gallop at RAL (jrg@uk.ac.rl.inf).
The current graphics standards for application program interfaces, GKS, GKS-3D and PHIGS, allow areas to be output with solid colour or regular patterns. (In the case of PHIGS and GKS-3D, the planar area can be expressed in 3D.) Visualising a complex set of data may generate a large and complex mesh of polygons that need to be shaded. With GKS, GKS-3D and PHIGS, many output primitives would need to be called and most of the mesh's intersection points would be repeated. Also GKS-3D and PHlGS would be unable to express the shading effects caused by even simple lighting simulations. The application would be unable to take advantage of today's more powerful workstations.
PHIGS PLUS is a proposed ISO work item (intending to lead to an ISO standard) that originates from the de facto work on PHIGS+ in the USA and extends the GKS/GKS3D/PHIGS family. In outline, the relationship between them can be shown as follows:
In PHIGS PLUS there are several new surface output primitives. Some of these are multi-faceted, a set of polygons which can be in different planes from each other. An example is the quadrilateral mesh, which is a rectangular mesh of quadrilaterals and is suitable as the basis of a carpet plot of functions of two variables. Other output primitives in PHIGS PLUS are parametric surfaces, an example of which is the non-uniform b-spline surface.
The other major addition in PHIGS PLUS is the simulation of lighting effects.
The importance of PHIGS PLUS is that, as well as being a potential ISO standard, it will be the target for future implementations of high-powered workstations.
The technical content of PHIGS PLUS is still fluid and RAL works, via BSI, on the responsible ISO working group. There are two main subjects, on which it would be useful to have comments from potential users:
Readers who believe that they are likely to start using PHIGS PLUS sometime over the next two years are invited to take this opportunity of influencing the standard.
Since the UNIX system first became widely available with the licensing of Version 7 of the operating system, three major variants have evolved: System V, Berkeley System Distribution (BSD), and XENIX. UNIX System V Release 4.0 (SVR4), from AT&T and Sun Microsystems, is intended to consolidate the features of these variants into one unified UNIX standard.
The BSD UNIX system has been the most popular among the scientific and engineering community, and forms the basis of the operating system of many technical workstations, such as SunOS. UNIX System V has been dominant in the business and commercial market, while XENIX systems are dominant in the smaller desk-top market.
The base operating system of SVR4 will provide portability by establishing a common computing environment across different vendors' systems, and will provide scalability because it will run on the entire range of computing systems from PCs to mainframes. For each architecture, an Applications Binary Interface (ABI) is defined which will enable application programmers to produce binary packages that will install and run on any ABI conforming computer.
All the familiar UNIX commands and functionality will be available in SVR4 though there will be some merging and pruning of commands and options. SVR4 will be written in C from a base which will be System V Release 3.2 but with all the major BSD facilities also available. The philosophy of SVR4 is to provide a menu of software, from which OEMs will select a subset to implement on their particular machine.
Basically, to the average UNIX user everything will appear much the same as before, except that there will now be more of everything (see separate article). It is at the lower levels of the system where the major changes will occur; things like memory management, shared libraries and STREAMS/sockets. The bottom line is that if a program worked under System V.3, or under XENIX, or under SunOS, or under 4.[23]BSD, it should work with SVR4.
SVR4 is intended to provide an ' open system' that will run on many architectures and be supported by many vendors. The advantage of an open system is that an application will run, without much modification, on each architecture. For software developers this should eliminate the need to produce several versions of their software, and for the users of applications it means they need only purchase one copy and be able to run it unchanged on a variety of machines.
There will be several features in SVR4 to assist the applications writer. Programmers who use the C language will have access to an ANSI C compiler. New user interfaces have been developed to help programmers maintain consistency in the 'look' and 'feel' of the software they produce. In addition, the Network File System (NFS) will be available for the transparent sharing of data across a network.
The schedule for SVR4 is that the source code will become available towards the end of 1989 and that ports to various manufacturers' machines will appear in the first quarter of 1990.
The separate article describes the principal features and functionality of SVR4 in more detail.
The Materials Group within the Cambridge University Engineering Department has, in addition to more conventional research, several projects involving a greater or lesser computing content. These activities fall under three headings:
Professor Mike Ashby heads the research group and is developing a methodology for materials selection at the conceptual stage of design. The method depends on a set of materials selection diagrams which show how different material types group with respect to particular pairs of properties, eg density and stiffness. A grant from the Leverhulme Trust has enabled the group to employ a programmer to create a front-end to a materials database using these techniques based on a prototype which has been running in the department for over a year. The database system is being written in C and is initially targeted at MS-DOS machines, but the database and graphical libraries are portable and versions for Unix workstations are planned.
Materials process and property modelling has a long history within the group, starting with Professor Ashby's materials deformation maps now widely used in materials engineering teaching. A series of PhD students and post-doctoral researchers have produced analytic models describing a number of processes in terms of the fundamental mechanisms operating within the material. Examples of these mechanisms are vacancy diffusion, dislocation glide, precipitate coarsening, heat flow etc. Successful models have covered creep deformation, sintering, hot isostatic pressing (HIP), weld-affected zone development, metal on metal wear and laser-transformation surface treatment. Current projects are active in ceramic wear, more complex hot isostatic pressing systems and the formation of surface-mount solder joints. In recent years these projects have programmed exclusively in Turbo Pascal on MS-DOS machines and this will largely continue. The solder project (in conjunction with the Loughborough University of Technology) will use commercial heat-flow finite element packages (courtesy of SDRC) on a Sun 3 to study realistic joint geometries. The original deformation mechanism map program has been rewritten in Turbo Pascal (from the original Fortran) and it is hoped to make it available to universities, for a small charge, during 1989.
Materials property data interchange is assuming increasing importance in computer-aided engineering and advanced manufacturing. Dr Philip Sargent, SERC Advanced Research Fellow, is chairman of the BSI advisory panel for this area and studies both the technical and organisational implications of these developments. This necessitates research in appropriate grammars for data interchange languages and data modelling of materials properties and test results. This work is currently done on MS-DOS machines using Icon, lex, yacc and the Lupine parser for the Express language from CADDETC (Leeds).
The eventual aim is to work towards integration of predictive materials modelling programs with materials databases in a safe and useful way.
At its meeting on 14 March, the Computing Facilities Technical Advisory Group (CFTAG) decided to invite bids to carry out a pilot study in the provision of public domain numerical software for use by Engineers.
The contract will be for a period of two years and dependent upon the results of the study further contracts may be awarded at a later date.
To establish a software archive and distribution centre for public domain numerical software for EASE.
Bids should be made to me at RAL by 15 May and will be considered by CFTAG at the 12 June meeting. The outcome of this will be relayed to those who bid by 19 June 1989.
I was very interested to read this report in the Engineering Computing Newsletter, Issue 10.
I was, however, very puzzled by the report on IKBS performance on the Sun4, suggesting that the Sun4/11O was slower than a Sun3/60 by a factor of 0.5 to 0.7.
This is totally inconsistent with our experience of Sun3 and Sun4 speeds, and I wonder whether the difference has to do with the particular implementation of Prolog tested, which was not specified?
We do not have a Sun4/100, but do have two Sun4/260s, both used simultaneously as file-servers and time-shared machines, for which they are excellent. On an admittedly small set of Prolog tests, running Poplog Prolog suggests the following approximate speed ratios:
Sun3/60 1.0 Sun 3/280 1.8 Sun4/260 3.7
Assuming that the Sun4/110 is about 0.7 times the speed of the Sun4/260, as claimed by Sun, this would give
Sun4/260 2.6
Even if the Sun4/110 were only 0.5 times the speed of the Sun4/260, the ratio would still be about 1.9, ie around three times the speed suggested by the report.
Is it possible that the IKBS tests used a compiler or interpreter that was not well tailored to a RISC architecture?
I should point out that we do not find the Sun4/260 is uniformly superior to the Sun3/280. Integer multiplication is slightly slower, since it all has to be done in software.
For most tasks we find the speed ratio of our Sun4s relative to the Sun3/280 is between about 1.5 and 2.5, with the average being about 2.0. This would make the Sun4/260 mostly between 2.6 and 4.5 times the speed of a Sun3/60. This is more consistent with your NON-IKBS assessments of the Sun4/110.
I thought it important to give you this information as your report would otherwise suggest to some AI users that a Sun4 would be far inferior to a Sun3.
The two Sun4/260 machines have the following configurations:
The second one happily supports at least 25 interactive Poplog users at the same time as serving 6 discless machines.
First, I would like to thank Aaron for writing. Input of this type from users is always welcome, and helps to ensure that the information we provide is properly understood. Benchmarking is always a difficult area to undertake, and is very easy to misinterpret.
Due to the limited objectives of the Assessment, it proved possible to run only one program representing IKBS-type work. The Sun4 was running under an early version of OS3.4 at the time (last August), and the IKBS test was compiled under a non-optimised version of the portable Edinburgh Prolog on an early version of the C compiler (that was all we had available). There was a wide spread of results from the repeated runs we made (as indicated in the table). Recent results with OS4.0.1 and a new C compiler have reduced this spread considerably (although there is still more variation than might be expected). This produces a new average result which shows an improvement over the previous figures by a factor of 2. The non-optimised compiler could well account for much of the remaining speed difference.
There have been a significant number of new announcements by several workstation suppliers since the Assessment was undertaken, which will significantly change the performance figures. We intend to repeat the tests later this year, and will publish up-to-date information (including Operating System version etc). Meanwhile, we would encourage those who have done their own tests on any of these machines to share them with the rest of the Community, via this Newsletter. Indeed any comments or opinions about any articles in the Newsletter are very welcome.
This workshop on Thursday 1 June at RAL is based round the Kent Software Tools, which is a suite of tools to help make programmers more productive. These tools run on Sun-3 (under Sunview or XII) and VAX workstations, and are issued free to EASE users. They are already in use at over 100 sites, and one of the tools recently won the BCS award for the best technical achievement in 1988.
The Kent Software Tools cover debugging, profiling (finding out, interactively, where a program spends its time), file comparison, manipulation of the file system, and production of documentation using hypertext techniques. The tools relate both to C and to Fortran programs. (There is also a special extra set of Fortran tools, based on TOOLPACK.)
The workshop has three purposes:
During the workshop, a set of workstations will be available for on-line demonstrations and discussions; attendees who bring cartridge tapes will be able to take a copy of the latest release of the tools.
Sun UK User Group 25 April 1989 R22 Lecture Theatre, RAL
Software Tools Workshop 1 June 1989 R22 Lecture Theatre, RAL
UIMS (User Interface Management System Workshop) 18/19 September 1989 UMIST
The Fifth Alvey Vision Conference (AVC 89) 25-28 September 1989 University of Reading
Your new Editor is Sheila Davidson. As an old-timer at RAL but new to the Engineering Computing scene, I shall be looking after your Newsletter now.
We are sorry to lose Fran Childs, who has also started a new post, and thank her for the first issues of the Newsletter.. We wish her well in the future.
Whether you are a contributor or a reader, I would encourage you to send articles and give me feedback, so that together we may provide useful lines of communication for the Community.
AT&T and Sun Microsystems have recently announced details of UNIX System V Release 4.0 (SVR4.0). This is their proposed new standard for the UNIX operating system which merges System V, BSD, and Xenix features into one unified version. SVR4.0 is being promoted by UNIX International, an industry association of companies and software developers, who include AT&T, Amdahl, Control Data, Data General, Fujitsu, ICL, Olivetti, Prime, Pyramid, Sun and Unisys.
A further group of companies have formed a rival conglomeration with the aim of developing a UNIX standard based on IBM's AIX implementation. This group is known as the Open Software Foundation and includes: DEC, IBM, Apollo, Nixdorf, Siemens and Hewlett-Packard. Talks between the two groups took place during 1988 but broke down in disagreement and at the moment users look like being faced with two different standards for the UNIX operating system. This article describes the principal features and functionality of SVR4.0 and is based on information provided at the Software Developer Conference held in London (Oct. 1988).
Three major variants of the UNIX system have evolved: BSD, XENIX and System V. UNIX SVR4.0 is the first consolidation and implementation of these variants and the first realisation of current standards. SVR4.0 encompasses the POSIX Pl003.1, ANSI C and SVID (issues 1,2 and 3) standards and is a complete superset of SVR3.2 and Microsoft XENIX. It also implements the majority of BSD and SunOs interfaces.
The SVR4.0 command set represents a unification (with pruning and selection) of the functionality found in SVR3.2, BSD and XENIX. Those commands that could not be merged into the SVR4.0 command set have been placed into a compatibility package. SVR4.0 derives much of its base functionality from SVR3.2, the major features of which include: the STREAMS character I/O mechanism; advisory and mandatory rue and record locking; semaphores, shared memory and message queues; synchronous rue writing; named pipes; full 8-bit characters.
SVR4.0 supports the System V shell, the C shell and the Kom shell. All three shells support SVR4.0 job control signals as defined by POSIX Pl003.l. It also supports rue renaming, memory mapped files, file synchronisation and rue truncation system calls.
The diagram below illustrates the major feature components of SVR4.0 and their derivation.
SVR4.0 implements a Virtual File System (VFS) concept to organize all file system operations. VFS is a merge of the System V File System Switch (FSS) and the SunOS VFS mechanism. VFS provides a file system type independent interface to programs and users. File systems that will be supported under SVR4.0 are SysV, RFS, NFS, the Fast File System from Berkeley and a new Process File System ( /proc). Both hard and symbolic links are supported.
SVR4.0 includes the STREAMS mechanism introduced in SVR3.0 plus termios from POSIX and new features such as STREAMS pipes and STREAMS tty. The STREAMS mechanism addresses the character I/O problem with traditional UNIX systems. For monitoring activity on multiple file descriptors simultaneously, SVR4.0 supports the BSD defined select system call as a library function on top of poll.
UNIX SVR4.0 supports a variety of interprocess communication (IPC) facilities including: message queues; named pipes; named Streams; pipes; Remote Procedure Call (RPC); semaphores; shared memory; and the Transport Layer Interface (TLI). The process scheduler in SVR4.0 now supports two process types: timesharing and real-time.
The ANSI X3J11 committee's draft standard for the C programming language is fully supported by the SVR4.0 C compilation system. In addition, the standard C libraries will include features defined by the IEEE POSIX PI003.1 standard.
SVR4.0 allows you to use ANSI C, but does not require it. UNIX SVR3.0 introduced static shared libraries in which object modules from libraries are no longer copied into the a.out file. SVR4.0 is compatible with existing executables built for static shared libraries, and supports dynamic linking for C programs. The implementation is based upon that in SunOS 4.0.
Release 4.0 extends the networking capabilities of UNIX System V to include the communication services defined by DARPA and supported by the sockets interface of BSD. Release 4.0 brings together RFS from System V and NFS from SunOS, including the capabilities for RPC, a high level network programming facility, and External Data Representation (XDR), a machine independent data representation capability.
UNIX SVR4.0 networking is founded on SYR3.0 networking facilities which provide: the uucp basic networking package; Remote File Sharing (RFS) capability; STREAMS mechanism, which supports modular development of networking protocols; Transport Level Interface (TLI), a protocol independent programming interface to transport protocols; and the network listener for connection management.
As part of the unification of the UNIX System, many of the features of the Berkeley Distributions 4.2 and 4.3 have been incorporated into UNIX SVR4.0. The sockets mechanism is widely used in BSD systems and derivative operating systems such as SunOS. In order to support the embedded base of sockets applications, UNIX SVR4.0 provides a full implementation of the sockets facility. The TCP/IP Internet package in SVR4.0 is a comprehensive implementation of the DARPA protocols and commands, and also includes the popular BSD r* commands e.g. rep, rlogin, rsh, rwho, In this release, commands are implemented over the Internet protocols, and will not operate over other networks. inetd is provided for server management.
SVR4.0 provides two distributed filesystem mechanisms: RFS introduced in SVR3.0 and designed for file sharing in a UNIX System V environment and NFS introduced in SunOS and designed for file sharing in an environment containing different types of operating systems. Both RFS and NFS have been integrated into SVR4.0 under the VFS facility. RFS permits the sharing of remote devices, but does not support diskless nodes. The NFS in SVR4.0 is Version 2, but Version 3 is being worked on with the aim of giving better heterogeneous support.
The uucp family of networking commands provides the base networking utilities in UNIX System V. SVR4.0 uucp supports higher speed networks such as LANs, and is based on the Release 3.2 HoneyDanBer implementation. This has many new features but still works with older versions of uucp.
User interface extensions in UNIX SVR4.0 provide services that simplify the development of applications that interact with users. Style guidelines and development tool kits are provided to help programmers maintain consistency in the look and feel" of the user interface. UNIX SVR4.0 includes extensive features and functionality in both the character-oriented and the graphic-oriented user interface.
The FMLI (Frame and Menu Language Interpreter) Toolkit provides a specific Look and Feel to character applications. FACE (Framed Access Command Environment) is an FMLI application which provides a friendly user interface to UNIX System V commands through menus and forms, and presents the end-user with a desktop environment for organising files, data and applications. The ETI (Extended Terminal Interface) Toolkit allows for customisation of the interface to character applications.
The User Interface on the graphical side is the OPEN LOOK Graphical User Interface (GUI), which is built on familiar concepts pioneered by and licensed from Xerox. This has a specific Look and Feel that gives all its applications a consistent appearance. Two toolkits are provided with SVR4.0. The OPEN LOOK X Toolkit is based upon the X Window System and the OPEN LOOK NDE (NeWS Development Environment) Toolkit is based upon the NeWS graphical platform. Both provides the necessary software for designing applications with the OPEN LOOK Look and Feel.
The window platform for UNIX SVR4.0 is an X11/NeWS merged window system. The merged server is completely compatible with applications developed for either the X11 or NeWS environments. Further, the X11/NeWS merge allows an application developer to mix window system calls within a single application. Below the merged server, SVR4.0 provides both the raster-op and stencil-paint image models.
UNIX SVR4.0 defines Application Binary Interfaces (ABI) for computers running UNIX System V. In addition to supporting a standard source-code interface across a range of processor architectures, UNIX SVR4.0 will also support binary compatibility within a number of micro-processor chip architectures including: Intel 80386; Motorola 68000 and 88000 families; SPARC; others - MIPS, Intergraph Clipper.
An ABI allows applications software to run interchangeably across the platforms that support the ABI as easily as software can run in the PC world. An ABI precisely defines the rules that govern binary software compatibility for UNIX systems that share the same architecture.
The structure of the ABI will consist of generic, architecture-independent components, plus architecture-specific components.
