Editorial: Is There Room for APL?

Transcribed from the printed issue; not yet reviewed. Read it in the PDF of the issue, page 3

As we move from the Eighties into the Nineties, it is interesting to speculate on the shape of the computer world ten years hence. It is also (just) passing the point where I can claim 10 years' experience of APL, so I suppose I could use the simplest of forecasting algorithms (beloved of the Foreign Exchange community): tomorrow will probably be a bit like yesterday.

In 1979 there were no IBM PC's, few interactive terminals (and most of those were the noisy typewriter sort), almost no colour, practically no graphics. Oh, and there was APL, complete with a shared-variable file system and the ∇ editor. APL did lots of amazing things that nothing else could do (like spreadsheets) and it quickly found a following in the business community. It also spread very slowly into the related worlds of mathematics, Operational Research, and Statistics; as it happens I first came across APL as a tool for building a distinctly complex production planning system: a typical piece of O.R. work.

In 1990, you can tote an IBM PS/2 around on the London Underground; this machine sports an elegant grey satchel, and packs nearly 4 times the punch of the entire CPU complex that Rowntrees owned in 1979. It has (among the spreadsheets and wordy processors and simulation modelling tools) an APL interpreter of significantly greater power than anything available in 1979, and it does rather good colour and graphics. In spite of this huge increase in power and functionality, there remains one rather unpalatable fact: the business community is deserting APL in droves, and at an ever-increasing rate.

So ... is there room in the hard commercial world of the Nineties for APL as a serious professional programming language? If not, where is the language likely to pop up next? Let me offer a few speculations, and maybe (if we keep going that long) the lucky editor of VECTOR 16.3 can have some fun taking them apart:

• the majority of big companies (like Rowntrees and British Airways) will take steps to isolate the use of APL into specialist areas like O.R. They will cease to use it for 'mainstream' systems development, and will progressively eliminate existing APL systems from strategic areas of the company. This process is likely to take longer than 10 years, but by 1999 I would expect to see almost no new APL systems in these areas, and many current flagships will have been decommissioned.
• specialist areas who are big enough to resist the drift (I reckon you need a department of at least 6 to support continued APL expertise) will continue to get the support of top management, and will continue to irritate I.T. professionals well into the next millenium. They will get off the mainframe or will be (gently but firmly) kicked off. PCs are already an attractive alternative for APL; they become that much more so when you need to work under cover!
• the slow spread of APL awareness in the mathematics community will continue, and will pick up significantly as the next generation of computers is made available to our schools in the mid-nineties. I suspect that when we look back on the development of APL from a long way ahead (say 2020) we may see a steady growth in the use of APL by the mathematicians, masked by the huge blip of its temporary foray into the world of commerce.
• APL will begin to be appreciated for its internationalism. A very large percentage of the world's mathematics takes place to the East of what we used to call the Iron Curtain. In the Soviet Union they at least use a familiar system of writing, although some of the symbols differ; in Japan even the system is different. I know that I can read Japanese APL as easily as I can read American APL or German APL; isn't it time someone told the Soviets about it?

I suppose it would be equally valid to ask 'is there room for Fortran?' or 'is there room for Cobol?'. If you look at commercial programming, you will again see a rapid decline in conventional languages, as the power and flexibility of packages converges on the requirements of the 'average' company. However I cannot see the world's physicists giving up Fortran, and there will always be the odd hole for Cobol code to nest in.

I hope and believe that APL is quite different; that it is in some peculiar way orthogonal to the Fortrans and Cobols and Cs of this world. However hard the I.T. professionals try to stamp us out, we APLers are always likely to pop out again somewhere quite unexpected; I am reminded of the biology of the Common Horsetail, which (according to Stephen Young in a recent New Scientist):

"Aeons of evolution have have left with an unparalleled degree of resistance. Throughout last summer our tarmac contractor returned to our drive - chlorate or Paraquat in one hand, sledgehammer in the other - at least half a dozen times. On each occasion he was routed by this most virulent of weeds.

Nothing short of a chemical holocaust will kill it. If you drench the visible stems with a systemic herbicide - one that spreads through any normal plant - the horsetail shrugs it off, refusing to fall for so simple a ploy. While the docks and dandelions die back, the horsetail comes back for more. In any case, its tissues are resistant to many commonly used herbicides, presumably because its biochemistry departs from that of the common herd."

A Happy New Decade to all you Equisetum arvense out there.