Who's OnDr. Greenes is one of the people who designed M at the Massachusetts General Hospital. Today we take over where we left in MWM-001. If you missed out last issue, click here.Who's On: Dr. Robert A. Greenes, MD, PhD
The DesignMUMPS was developed to handle / cater for:
The goal initially was to achieve efficient performance on small time-shared minicomputers by having a terse language so that program segments could reside in small partitions in memory. Traditional time sharing systems, at the time, required considerable delay to swap programs in and out whenever a time-share context switch occurred. In CPU-intensive jobs that was not much of a problem, but in applications which were dependent on a lot of user input/output transactions, the time required for a context switch could be very large in relation to the amount of processing actually required. Thus it was beneficial to have many small partitions in memory ready to be run at all times. Much of the latter constraint disappeared over the years as memory became plentiful and cheap. It was determined that an interpretative language with terse commands, string handling, and hierarchical sparse data structures for transient and persistent storage would address these problems. An integrated time-sharing executive, I/O handler, and interpreter were implemented on a PDP-7 using assembly language. By 1969, MUMPS was implemented on a PDP-9 with 24,000 words of 18 bit memory (approximately 53KB) and a 3MB Burroughs fixed head disk. The system supported 16 concurrent users. A typical IBM compatible computer today is sold with 16,000 KB and has a hard disk greater than 1,000MB. MUMPS is a distant ancestor of which is JOSS developed at the Rand Corporation in 1964. It was influenced by languages such as TELCOMP and STRINGCOMP developed by Bolt, Beranek and Newman, Inc. and FILECOMP (General Electric Corp.). The re-entrant interpreture, time-sharing monitor, I/O monitor and buffers took up about 8,000 words of memory. This is one third of the available memory on the computer and corresponds to about 18KB of today memory.
MUMPS was here to stay, but…In the early ‘70's there were already starting to be different dialects of MUMPS, so the MDC was established to foster standardization. The main outlier has always been MEDITECH (Pappalardo's company), with its MIIS and Magic languages, and its unwillingness to foster the standard. The Future"I have not followed how M has evolved and is being handled in recent years, but feel that the main problem with MUMPS for decades, which I have pointed out in previous papers, is the tendency of the MUMPS culture to remain somewhat insular, and to be slow in adopting open standards, integration with other third and fourth generation tools, ease of passing data structures in/out of MUMPS, and becoming more user-friendly. Much of the original terseness and limitations of the language, furthermore, which was designed for I/O partitioning efficiency, are unnecessary in view of the cheapness of memory, the availability of pre-compilers and other means for providing more modern language features. I think MUMPS/M will recede to the back-end of systems, since it cannot match the GUI capabilities of modern toolkits designed for that purpose. It will continue to be useful to the extent that its flexible data structures are desirable, and to the extent that data can be passed in/out of M to other languages. Also it will depend on the extent to which relation DB's are able to achieve sufficient efficiency to replace B-tree data models of M. Another serious limitation of M is the lack of explicit DB schemas, which limit participation in open systems and distributed operation with other environments." "An advantage of this approach … has been the increased flexibility that can be built into a system rather than designed around. At the same time, both the flexibility required for research and development, and the efficiency required for service operation can be achieved with a computer whose size and cost are well matched to the requirements of the problem area."Excerpt from the concluding paragraph of the paper "A system for clinical data management" by R.A Greenes, A.N. Pappalardo, C.W. Marble and G.O. Barnett, presented at the Fall Joint Computer Conference, Las Vegas, NV, 1969.We would like to thank Dr. Robert Greenes for the interview and for the papers:
|
" " ± Can you be of help?Many organizations produce promotional stuff to give to clients. MWM needs goods to give away in reader competitions and other activities run on MWM.If you would like to donate something for these projects, our readers will be utterly grateful and we'll make it a point to list your company in our Thank You page. If you would like to know more (or maybe even sponsor your own companies's specific trivia, questionnaire, or game) why not talk to us. Our e-mail addresses are mwm@mcenter.com and chribonn@softhome.net. Thanks for your support C O R R n d t h e W i n n e r |
Dr. Greenes is
the director of a medical informatics research and development
laboratory at Harvard Medical School, located in the Brigham and
Women's Hospital which he established in 1978. This laboratory is
known as the 