Aucbvax.2453 fa.works utzoo!decvax!ucbvax!works Mon Jul 27 06:54:21 1981 Collected responses on extensibility and design >From WorkS-REQUEST@MIT-AI Mon Jul 27 06:49:14 1981 WorkS collected responses on extensibility and design ---------------------------------------------------------------------- Date: 26 Jul 1981 20:36:10 EDT (Sunday) From: Bernie Cosell Subject: mundane systems I must be missing something. It seems that `mundane' means something a little different from what I thought... It looks like there is a confusion between `under designed' and `under powerful'. I originally thought `mundane' to have been intended to connote `under designed', but it seems that most of the comments about it immediately make arguments based on a connotation of `under powerful'. For example, in reference to Jan's car-shift image, I wonder what her choice would be if it were between the grubby automatic and a multi-hundred gear shifter with hidden catches and missteps around every corner: miss first gear and the windshield wipers turn on, fumble a double clutch and your brake fluid all leaks out (good luck stopping). And.. the way the gearshift has been designed, it is almost impossible to find a comfortable subsets of the gears that you can use - even with lots of practice you still occasionally bollix something simple up and have something truly perplexing happen. And... the way you get anything fixed is to be given a tool box and a service manual and be told you are on your own (you don't like having ten different reverses? just get in there and change it). Although I use EMACS and am fairly proficient at making it stand on its head, I resent the amount of irrelevant expertise I had to master to be able to move about in that world, and the need for introspection at every level: far far too many things were clearly under-human-engineered (they may have been OK for programmers, but certainly not for people). [For example, I believe that any sensible design would have its commands mostly ordered by danger and globality and one would try very hard to make the very easily typed commands be only the local, non-dangerous ones and make it progressively harder to `get at' the more wide ranging, damaging commands. Having `delete a character' be next- door to `muck with the whole region' (most of the time when you hit it by mistake, you probably won't even know where you last left the mark). In my (painfully crafted) EMACS environment, the control and meta functions are mostly only those mostly useful in on the fly correction of running text (delete a word, twiddle chars, etc.), my VT100 keypad has the global/dangerous commands that I use regularly but that I a) want to be sure of not hitting by mistake, and b) usually appropriately want to pause a moment to make sure that I mean what I am about to type (delete region, undelete region, filter region, etc.). It aint perfect (I am hardly qualified as a human factors expert), but it is a darn sight less hostile and punishing to use than `naked' EMACS (and yes, my profile really does go and un-bind all of those dozens of silly functions hanging off of those silly keys).] I would have much preferred that someone with some taste and perspective had gone and made a whole coherent `environment' be available and as much as possible hide the other stuff. Given the choice, I think I would always go for the well thought out mostly useable tool - even if I can't personally `fine tune' it -- over the ill-thought-out facility that, due to its lack of good design, FORCES me to go mucking with it. /Bernie ------------------------------ Date: 26 Jul 1981 13:33:42-PDT From: SomeoneOnUUCP at Berkeley Although I agree with Newcomer's comments on easily-usable program, I am not entirely convinced that anything can be done about it. We can make simple models of editors, formatters, and automated desks -- and these are likely to be usable for simple tasks. But the real world tends towards far more complexity than we would like, and sometimes the only way to deal with it is to add complexity to our programs. Worse yet, the real world is really a plurality of worlds. I might be able to write a text formatter that satisfied my requirements perfectly, but would be perfectly useless in meeting the graduate school's obscure rules for dissertations -- rules like where the page numbers have to go, what pages should be numbered, when to use Roman numbers -- all manner of nonsense. And there's not a thing I can do about it, because such rules were (apparently) carved in stone 3000 years ago. This is not to say we shouldn't keep trying, of course. And one of best things we can do is to DOCUMENT our unstated assumptions and design goals, as these are often the best guide to the overall structure of a system. I found it very difficult to learn Berkeley's 'vi' editor, simply because such a statement was missing, and they had added so much syntactic sugar as to hide the basic structure. [ There was nothing with this message to actually identify the author. On behalf of everyone, I would like to ask Berkeley recipients to take the time to identify themselves within the text of the message when your mail system itself will not do it for you. A simple way to do this is to "sign" the message at the end as many ARPA people and a few UUCP users do already. Without some identification we are prevented from getting any replies back to you at all. And it becomes very difficult to discuss anything when we cannot figure out who is saying what. Thank you. -- RDD ] ------------------------------ End of collected responses on extensibility and design ****************************************************** ----------------------------------------------------------------- gopher://quux.org/ conversion by John Goerzen of http://communication.ucsd.edu/A-News/ This Usenet Oldnews Archive article may be copied and distributed freely, provided: 1. There is no money collected for the text(s) of the articles. 2. The following notice remains appended to each copy: The Usenet Oldnews Archive: Compilation Copyright (C) 1981, 1996 Bruce Jones, Henry Spencer, David Wiseman.