During this quarter, the Group lost its two sandwich-students, Alan Pitchers and Chris Hims, who both returned to college after a 12-month and 7-month stay respectively.
Both made very worthwhile contributions to the Group's achievements, Chris Hims doing some valuable work on SYSTEM JOURNAL, performance measurement and on the benchmark scheme, while Alan Pitchers developed the Atlas logo sequence to be used on the Laboratory's films and produced a number of utilities for the SPROGS system.
We have gained only one student in replacement, Ahcene Idris-Bey from Portsmouth Polytechnic, who arrived on 8 July for a year. David Duce joined the Group on 30 September to work on the 1906A system and John Rushby joined at the start of October to work on graphics.
At the end of the quarter, Anton Walter moved from the 1906A side to the Graphics Section work on SPROGS on the 360/195.
The Group structure now is:
Again it has to be reported that GEORGE 4 Mark 8 is not yet imminent. With holidays and various conferences, this has been a quiet quarter, particularly from the point of view of new facilities for users.
The reliability of GEORGE Mark 7 has improved and a limit of one or two breaks a week seems to be achievable. Even then, many breaks are due to some new aspect of GEORGE being explored but the faults are fairly well understood.
The measurement of GEORGE has shown good progress and is well on the way to some sort of conclusion. However, it appears that it will always be necessary to maintain effort in this direction.
Development in TASK and its associated compilers has been mainly to make use of the block read facility which reduces the number of excursions into Executive and GEORGE for input to compilers. This appears to reduce compile times by 25% or so.
NUTS has been developing steadily throughout the period. It is hoped to provide more effort on this project as the poor utilities are a weak feature of the system.
The source and binary of GEORGE 4 Mark 8 has been received. In discussions with other 1906A sites it appears that Mark 8 might be adopted at Christmas or the New Year. There seems little enthusiasm for moving to Mark 8 considering the satisfactory state of Mark 7.
The cross hairs on the Tektronix can now be read by an on-line program, using a Tape Reader stream.
Mends will be needed for the communication front end project to allow the front end to control properties, associations and attributes. This should allow devices to be assembled together to form an RJE and to be freed as an RJE logs out. Currently, problems exist with regard to allowing output to accumulate for a property which has no devices.
A program is being written using the NUTS system to edit a dump tape. In this way many of the directories can be removed to produce a small filestore with a short BACKMAP time. This will facilitate manipulation of filestore discs, etc, which has to be done outside the filestore although it is more convenient to do this under GEORGE.
The Oxford University Secure Jobwell has been introduced. This required amendments to remove errors (concerned with RJE stations) and to add features to smooth its introduction. GEORGE was also mended to allow the EXPRESS command to be used for non-started jobs. High Level Scheduler was re-written and there remains further development work to do on this.
The technical problems were less severe than the problem of change to the user interface, especially as there are initially no advantages apparent to the user. However, it was thought better to introduce the new interface during a period of low machine activity (September) rather than to wait for more facilities but perhaps being forced into a hasty implementation during periods of heavy prime shift load. (The old JOBLIMIT of 90 was achieved several times earlier this year.) The Oxford Secure Jobwell is limited but will have the same interface as the ICL Secure Jobwell which is expected early next year.
The bugs now seem to be all removed and the Secure Jobwell is regarded as permanent. The machine loading since its introduction has been too low to make any assessment on efficiency.
With the introduction of the Secure Jobwell it was necessary to provide users with a method of tracing the progress of their jobs. The contents of :SYSTEM.JOBLIST are converted to a readable form which is edited by the STATE macro to list jobs with given jobnumbers, jobnames or usernames (SRP).
Responsibility for logging weekly performance statistics has been handed over to Operations Branch.
The jump in efficiency observed in May due to the removal of an Executive mend (see below) has been maintained and record figures have been achieved during the period.
The benchmark test to prove the bad Executive mend was an outstanding success; the test ran almost 50% longer on the old Executive (Technical Notice 87). The benchmark was also used to cost the saving of block reads in the PLASYD compiler (Technical Notice 88) and to measure the cost of a null job in GEORGE (less than 1 second).
The work has proved the value use of having a standard benchmark and its in costing aspects of the system. Unfortunately, it has barely scratched the surface of its possible revelations and further work may have to await more staff.
The new method of SYSTEM JOURNAL archiving is now in use and runs automatically when the size of the unarchived files is above a certain level. Documentation is currently being produced.
A program has been written to analyse the compacted SYSTEM JOURNAL archive and produce a distribution of numbers of jobs with various jobtimes and their mean terminal times (AI-B).
A new JOURNAL Listing Program (called JLP) has been provided to replace the ICL version. The program uses the ALGOL68 JOURNAL interface, it accepts date, time and duration parameters for the section of JOURNAL to be read and will list either selected messages or all but certain messages.
TASK now records the progress of each job run in SYSTEM JOURNAL. Programs have been written to list this information in a readable form, and to produce statistics of the use of TASK over a given period.
The filestore accounting scheme was further enhanced and tidied early in the quarter. The automatic production of LSTR listings for usernames at the time of their losing CONTEXTA privilege was implemented in the manually-started control system. Use of the control system was re-started by Operations Branch after a short pause during change-over of systems. Evidence of the usefulness of the system was seen in an immediate reversal of the previous four week growth of filestore size. The growth had been in approximately unit steps per week, and the subsequent recession after re-starting control was in similar size unit steps for a few weeks. A new format output of users' filestore control information, which had been planned for some time, was implemented, and this has eased the task of operating the system as the new format ties in closer with that of the notes sent to users. Checks have been built into the system to stop unnecessary or incorrect re-runs of time-consuming or sensitive phases, particularly on restart after a machine failure. Start-up of the job has been placed under the DO macro, and the whole responsibility for running and controlling filestore accounting has been passed to Operations Branch.
The faults that were occurring in the accounts database interface late in the last quarter were finally traced to a complicity between the central database interface and the structure of a newly-generated database, and a mend to the relevant software allowed re-tracing of steps back to a correct database structure quickly followed by mending of the accounts fields and resumption of fault-free accounting.
Plans have been laid and work started on the re-structuring of the central accounting software. Two large benefits will be easy change-over to periodic JOURNAL accounting if required and much greater resilience of the system on machine breaks or other contingencies, removing the need for special manual intervention if anything goes wrong. Part of the new system will use two new budget extracodes for accessing DICTIONARY and some preliminary testing has been done on these. One run showed a bit dropped in one word transferred into DICTIONARY, but the event has proved non-reproducable; a close watch will be kept on this situation. Certain documentation relating to the new extracodes was found incorrect and different to that from other sources. However, investigation of the related GEORGE source clarified the correct state and a small mend is now planned to enable certain accounting functions to run other than under :MANAGER in the interests of security and simplicity.
Work has begun on a new editor for the accounts database (AI-B). The commands will be analysed by a compiler produced by the TREE-META system and be translated into FORTRAN calls.
As received from ICL, this had several undesirable features and some bugs. These were removed and the system considerably enhanced and operator documentation written. There is still one outstanding bug in that JUGGERNAUT does not recognise files currently on-line in its calculation of the filestore size and can overfill the filestore. This is currently being investigated by the author (P Tucker - ICL). It is planned to add block reads using the DCT PLASYD I/O package if time permits.
The advent of Secure Jobwell caused a major headache for TASK. Previous to this, TASK had ignored all store requests (via MZ parameter) for < 100000 words and was designed to operate in this store size. With control of this parameter now being vested in GEORGE it was necessary to modify TASK to run in any store size and to give an. appropriate error message if this was too small. The macro was also enhanced to set up the correct JD parameter for MOP-initiated TASK calls and also to cater for several more unusual user errors. The consolidation of large overlay binaries was enhanced to recover from too small a core value and finish consolidation if there was enough store available. Secure Jobwell also required a change in the JT parameter. It is now no longer needed for a large compilation. time as this is set by default to IS minutes.
Other enhancements include semi-compiled files being assigned EMPTY to avoid retrieval and the parameter APPCOMP to override this feature. The *MOP parameter has been changed to MOPEJ to conform to other systems. The maximum number of # peripheral parameters is now 25. A special property *A is now available to direct output to the central lineprinters and thence to the Program Advisory Office for remote users needing advice.
Several new versions of the PLASYD and FORTRAN compilers have been introduced during the quarter and it is planned to introduce a FORTRAN compiler using block reading of source files during the next quarter along with the ACL version of the XPCH consolidator. TASK now outputs its statistics to :SYSTEM.JOURNAL.
An enhancement to TASK planned for this quarter is the parameter LOG which will cause TASK to write information to a file about how a users' job ran so that he can easily access this information. A new version of the TASK manual is also under preparation.
TASK has achieved its first export success and is to be sent to Brunswick University, West Germany, for use on its 1906S.
Magnetic tape facilities have been added to NUTS during this quarter. They re-introduce the ATLAS concept of numbers for tape blocks and allow users and utility writers to address the tape on this basis.
Facilities to cater for parity errors, various types of end tape indicators, and detection and identification of ICL sentinels are included along with automatic implementation of the correct method of writing after reading.
NUTS is now able to COPY, COMPARE and PRINT standard ICL tapes. Similar facilities for non-standard tapes are implemented but, as yet, untried. NUTS can also compare disc files. New documentation in :NEWS.NUTS-DOC is to be issued shortly.
A new feature is the RUN parameter where a user can write his own utility to use the NUTS kernel routines and hence gain all the facilities they offer. This should make the writing of utilities a lot simpler and this is currently being evaluated by SRP. Documentation may appear in the next quarter.
COPYIN and COPYOUT were being implemented by D Parkinson (ICL) but he departed before completing them and DAD is taking these over.
The release of NUTS to external users awaits documentation and this is unlikely to be available for some while due to the pressure of current projects.
The AMED disc-file editor has now been released for general use. This is an editing program which supports the commands used by the ICL disc-file editor, XMED, and the UNIVAC editor used by the XRAY program as subsets of its own commands. A large factor of efficiency is gained over these programs by reducing entries to GEORGE, for example by using 1024-word *DA transfers and by using the block file reader facility on filestore files. AMED may be used via the DISCEDIT macro, or via NUTS.
The TRIM routine is used to reduce an AMED format direct access file to the minimum size that will contain the information in it. Versions of the routine have been included in the consolidator XPCH, the editor AMED and the PLASYD library (JDT). It has also been supplied for inclusion in the PLASYD and FORTRAN compilers (DCT) and the TASK and NUTS systems (GWR).
The work in hand on the PLASYD compiler mentioned in the last Quarterly Progress Report has now been completed: it is not intended to perform any further development of this compiler.
Not much has been done to GERONIMO this quarter. Operator requests addressed to RJE stations are no longer displayed, and the WELL command has been provided to allow the contents of the Secure Jobwell to be examined. This latter may be enhanced as the Jobwell progresses.
The file SUBGROUPS-PL has been created on the libraries disc 26. This is the semicompiled library used by the PLASYD compiler, and contains the JLIB routines (JDT), a set of input/output routines similar to those used in 1900 ALGOL 60 (DCT), and the monitoring routines from the PLAN library.
New input/output routines have been added to the PLASYD library, which use file readers and file writers. The input routines use file reader block reads, and provide a file-type independent input package for reading any type of serial filestore file. The output package uses a file writer and will become completely file-type-independent when the necessary GEORGE changes, currently in progress, have been completed. These output routines will be modified to use block writers when this facility becomes available.
New versions of these compilers are currently being tested, which incorporate the following features:
These compilers will be available as soon as they have been debugged: it is anticipated that they will produce a marked increase in efficiency in the 1906A when performing FORTRAN compilations.
A large amount of time has been spent on trying to implement the FORTRAN-to-FORTRAN optimizing compiler. A number of versions have been tried out, and the optimization procedures do not seem sufficiently robust for general use. A recent attempt to use the cross-reference routines only as an adjunct to FORTIDY have been hampered by corruption of the main and back-up disc copies of the modified source program.
Any further work on this project would require about a month of preliminary work in altering the original source again. As the resulting program is likely to be very big, around 200-300K, and also very slow, it has been decided not to carry on with this project for the present.
The consolidator has been carefully examined. Various proposals for enhancing the consolidation process have been made, but a review of TASK statistics suggests that the amount of time spent in consolidation is so small that the effort required to implement enhancements would be greater than the benefits.
A PLAN macro facility has been written. It is a pre-processor with card input and output and provides the following facilities.
This program was originally written as an aid to examining the consolidator. It is not documented or fully debugged.
Work is just beginning on examining the PLASYD compiler to see if a suitable pre-processor or built-in macro facility can be provided.
Two minor pieces of work are:
Several new facilities have been added to SPROGS this quarter and some internal changes have been made to increase the speed of the package:
The main effort has been on writing the programs for each computer that will send data both ways across the link. Testing has been confined to slack times on the 1906A up until now, because faults in the routines which deal with the basic handshake are liable to affect the rest of the system. When these basic routines are sorted out, testing can be more frequent. The PDP15 will, via the software being written, be able to communicate with a reading file and writing file in the 1906A. The PDP15 user can specify:
The BSILIST software can now send characters to the 1906A tape punch, as well as to the lineprinter and the filestore.
Version 1 of the PDP15 Interactive Graphics System is now complete and documented. A preliminary manual will be released within the next fortnight giving system details.
PIGS is a set of procedures which greatly simplifies the design of interactive graphics programs on the PDP15. The system is composed of an interpretive table-building language, MENDEL (for MENU DEFINITION LANGUAGE) and a library of procedures forming the PIGS run-time environment (PRTE).
In designing a graphics application program, MENDEL is used to describe the organization of operator commands into groups called menus. PRTE uses tables output by MENDEL to display menus of lightbuttons on the CRT, thus suggesting contexts of available commands. The operator may select a command using the keyboard, lightpen, sparkpen, or pushbuttons. He may schedule commands using the real-time clock. Once a command has been selected and its arguments defined, PRTE passes control to a pre-specified application procedure to perform the desired function. A new menu of commands may then be activated and displayed on the CRT.
As a test of the PIGS system and to uncover problems in it before WDS leaves, the SD4020 tape viewing program VIEW is being altered to run under PIGS. The opportunity will be taken to add some facilities, since PIGS allows new commands to be easily added.
As a consequence of changing to PIGS, the user will direct VIEW from the display keyboard or lightpen and most of the frequently occurring messages will be displayed rather than typed. Changes of operating mode (fast/slow search, printing mode) will be signalled from pushbuttons where possible.
The necessary rewrite includes changing graphics subroutine calls to use the FOG package, removing the command decoding section, altering input/output slots to avoid those used by PIGS. This is currently being tested.
The DECUS-supplied overlay builder, CHAIN, has been modified to allow the overlay specification to be input from a disc file and the load map to be output to a disc file. This should save a considerable amount of repetitive typing when debugging large programs. An auxiliary library scan option was also added. The new overlay builder may be loaded from DOS using the command, XCHAIN. Further details are available in PDP15 User Notes 2 and 3.
The Tektronix routines (Tony Burraston) are complete and documented in the PDP15 FORTRAN library, so the Tektronix 4010 can now be used for graphics, although currently only at 110 baud. An increase in line speed to 2400 baud is being considered, which will make the Tektronix a useful second graphics terminal under RSX. A preliminary version of the DMAC monitoring program (Gordon Dawson) is working and is being documented.
Work this quarter has been concentrated on providing tasks which will allow two users to access the PDP15 in a sensible manner. The supported version was still not available, but a large box containing a supported RSX and an unsupported DOS (!) has just been delivered.
The modified INSTALL and REMOVE have enabled tasks to be constructed without having to SAVE the system at the MCR console. A directory system has been set up to allow a user to access files in a DOS directory other than 'RSX', by giving access to a LUN-directory table within RSX Executive which exists already. Following the modifications to DOS CHAIN by WDS, similar changes were made to the RSX equivalent - TKB - which can now accept input from a file and send output to a file. It is perhaps interesting to note that the required changes were very much easier under RSX.
Under RSX, it is all too easy to lose disc blocks if the system crashes. Consequently, a disc checking program has been written to follow all known chains of blocks, such as directory files, DOS system files, RSX installed tasks, etc. Such a program can only be of informative use, since RSX allows users to obtain blocks for which no directory entries are made (apart from the setting of an in use bit). This program has already proved very useful in debugging system tasks.
As a prelude to the general conversion of standard DOS programs to run under RSX, an attempt has been made to implement VIEW. This implementation is proceeding in parallel with, but along different lines to that using PIGS, reported above. It has entailed the implementation of the magnetic tape handler and the (unsupported) graphics package, and has shown up some bugs in the latter. Changing from console switches to pushbuttons has enabled VIEW to be installed as a NORMAL, relocatable task. The program has worked successfully, but further changes are needed to speed it up, since there are considerably more overheads reading pushbuttons via Executive than directly accessing the console switches (forbidden to NORMAL programs).
In theory, it should be possible to run NORMAL programs in the upper 32K of core. However, the system hangs when these, or even much simpler non-graphics, non-tape programs are used. This looks like a timing problem (the cores are different speeds) and will require attention from DEC.
The soundtrack for FRAH's Hash Table film has now been completed. The PDP15 was used to control the synthesizer, the output of which was recorded on a sprocketed tape recorder by Peter Hadingham of Swift. Film Productions. This soundtrack was added as an optical soundtrack to a print of the film which was shown at IFIP74 in Stockholm and DECUS 74 in Zurich where it was favourably received. A paper describing the synthesizer interface was also written for the DECUS Symposium.
A new advertisement for insertion in Computer Weekly, etc, for graphics programmers has been designed. The copy features some output from the SD4020 to emphasize the graphical nature of the work.
Now that the new Colloquium is available, some experiments with the positioning of the loudspeakers have been performed. However, due to the problems with the white boards, special brackets will need to be made up to mount the loudspeakers above and in front of the white boards. The quality of the sound from the new equipment in the Colloquium was found to be very good.
Four polymer chain films, complete with introduction and final result sequence for each, were completed for Professor Edwards. These films were shown to him and were followed by a discussion between ACD, RET and Professor Edwards. The outcome of this was a request for a 5-minute amalgam film of the four films - and no further graphical work for the moment, as it was incidental to the polymer chain dynamics theory. Instead, Professor Edwards would like further work done on the theoretic side, in particular on the steady state of the chains, and the length of time it takes to unknot a knotted chain. Due to the current shortage of staff, no further work has been done, apart from the production of the 5-minute film.
The final programme for this Seminar, to be held on 31 October, has been settled and invitations sent to those interested. To date, there have been over 30 replies. It is hoped to follow this with a more general seminar next year.
This quarter has seen considerable activity in preparing a case for the purchase of an FR80 microfilm recorder to replace the SD4020. The FR80 is something of a Rolls Royce among plotters, having a 16K x 16K plotting area, ]6 mm, 35 mm, microfiche and 12in hardcopy cameras, colour, 256 grey levels and a small computer. The efforts have proved successful and ACL are now negotiating a contract with III for delivery in the New Year. The advent of this device will demand considerable software effort in altering existing graphics packages and tuning the FR80 software to suit ACL's needs. Considerable thought will also be required in deciding how best to operate it.
Manuals for the FR80 have been acquired, and an FR80 tape obtained from the Post Office at Leeds, who have the only other similar machine in the country. Programs have been written to decode the tape, and an FR80 interface is being written for SPROGS (due to its device independence, SPROGS will be the easiest package to change).
Progress has been slower than expected due to staff having to spend time on other projects, in particular the block read in compilers and the Secure Jobwell and Mark 8. Involvement in other projects is likely to be less in future but it is difficult to remove it completely.
The hardware has had mixed fortunes. A wiring fault on the BMC has caused trouble with the communications interfaces. However, there is still a suspected fault which manifests itself by the loss of the first two of the pair of SYNC characters at the head of a transmission. GEC do not appear to have any solution yet. Full acceptance of the machine and full payment await the solution of this problem. The CPU has proved very reliable. The power supply of both the core store, and to a lesser extent the CPU, tend to trip-out occasionally and no fault has yet been found. The paper tape reader has been a constant source of error and it has to be adjusted fairly frequently.
The 1906A /4080 interface is now complete and in the final stages of benchtesting. In the next quarter it will be tested with the 4080 and finally with the 1906A. This will hopefully be completed in the coming quarter.
In accordance with a request from SRC to bring forward spending, a case has been made for a disc, 32K of store, a duplex communication interface for EPSS, a 96-character chain for the 2050 and a CRC board.
In addition, cheap printers are being examined with a view to replacing the current 4080 lineprinter when the 2050 has eventually to be put in the field.
The new 1906A console has been delivered, rejected, rebuilt and redelivered and now appears satisfactory. It should be in during the next 10 days. Such devices as this (1200 bits/see printer and keyboard) have a future as MOP terminals. Several problems are apparent; first, the lack of a break-in, second, the fact that the buffer in the printer sometimes overflows. As was mentioned earlier, other such devices are being examined. Testing of the console has been done on the 4080.
Cassette tape devices are now being examined as a user has a requirement to connect one to a 2050 for data collection.
The only modification made to the Emulator this quarter was for the correction of a protocol error concerning RJE console messages to the 1906A. This error caused a 20-second delay in the reply from the 1906A, and now the response time is between 2 and 5 seconds.
Sussex have added a paper tape reader to their 2050 and after a few initial problems, tape data is being successfully transmitted to the 1906A.
The dumping facility added to this system has been used extensively to examine the data transmitted between the 2050 and the 360/195. From this a more complete understanding of HASP protocol has been obtained.
The 4080 synchronous communication has been extensively tested. A double buffering technique has proved to be very effective. This allows data to be read into a buffer and, when the buffer is full, the hardware automatically continues with the next buffer. Various methods have been tested to scan the input buffer as the data is received. This is necessary as the length of the transfer is unknown until the terminating characters are detected. The latest method is driven by timer interrupts and includes reading the Way Control Block to ascertain the number of bytes that have been received so far.
By using these and other techniques, communication has been successfully performed between the 4080 and the following systems:
The last two have been combined to achieve communication from the 2050 to the 360/195 via the 4080. The 4080 program simply receives a record from the 2050, sends it to the 360/195, receives the reply from the 360/195 and passes it back to the 2050.
To obtain the length of each record, the type of the record has to be investigated (ACK, NAK, ENQ or data). The program will also time-out if a record is not in a specified time and will take appropriate action.
Initial tests have shown that this method does not degrade the performance of the RJE station. This is because the 360/195 and HASP protocol can handle data faster than the 2050 peripherals.
To enable the GEC disc operating system to be used, an attempt is being made to utilise the SIGMA2 RAD (or ATLAS disc) over a communications line. Currently, a suitable disc driver has been produced, a bootstrap loader has been produced and an attempt has been made to dump part of the RL's 4080 disc onto tape, put this on the SIGMA disc and bootstrap the operating system in from it. Only minor bugs have so far prevented success. The last phase is to produce a disc operating system on the RL disc containing the specially-written driver and then transfer this by paper tape to the ACL 4080. It appears that there are no major problems now preventing this.
Considerable effort has gone into study of the core operating system. This is now well understood and has been tailored to the initial work. In particular, access to the synchronous interfaces and 1906A interface has been included (see above). Effort has also gone into providing a Teletype driver. This should allow multi-use of the 4080 for development which should prove useful.
A driver for an asynchronous communications board has been written (DCT), which successfully drives character devices at speeds between 110 and 4.8K baud: it is possible for example, to drive a Tektronix terminal at 4.8K baud, and provide break-in during output. With slight modification, this driver will control buffered devices such as a Tally printer, or the Dacoll printer to be used as the operators' console on the 1906A.
The basic structure has not changed. This will consist of a central KERNEL which will act as an exchange for all data and also contain the intelligence of the system. Communicating with the KERNEL will be several line drivers including the 7903 Emulator and the HASP protocol line driver.
It is expected that the 7903 Emulator will undoubtedly be the major task.
A number of minor projects have been undertaken in order to gain familiarity with the 4080. These include the implementation of a list processing package, a program to time certain facilities of the operating system, and a program to convert core dumps into a legible format similar to that produced by the GEORGE PRINT command (JDT).
It is hoped to base the core space allocator on the module for this purpose provided by GEC. This requires the source code to be made available by GEC, who are not inclined to co-operate very readily.
ACL has now obtained a 2.4 Kbaud port on the Manchester EPSS exchange. Initially this will be used for examining Teletypes on RJE access to the 1906A via EPSS. This may well develop into the major means of access to the 1906A. In this case, a wider baud link will be needed in due course. This project is not likely to mature for 3 years. The EPSS exchange opens in about one year. The EPSS connections have already resulted in useful discussion with other sites.
Visit to USA and Canada, J R Gallop
Computer animation used as an aid in teaching computer science (for IFIP74 Proceedings), F R A Hopgood
Computer Controlled Sound Generation (for DECUS 74 Proceedings), A H Francis
Some Aspects of the Production of Computer Generated Movies for Chemistry Teaching (for Proceedings of International Conference on Computers in Chemical Research and Education - July 1973). F R A Hopgood (with P Groves, Univ of Aston).
