It's 10:30 PM, and Weather Underground is showing "Today's" low temperature of 43 degrees F, because that's the temperature forecast for 6am tomorrow morning. Menawhile "Yesterday's" low is said to be 31.8. But in the wee hours of "this" morning, which would have been "Today's" low at this time yesterday, it was actually 41 degrees. Nowhere in the "Today" or "Yesterday" is the temperature it actually reached in the wee hours this morning shown, because it's too late to be "Today" (meaning until tomorrow AM) and too early to be "Yesterday" (because, ok, it isn't really Yesterday yet, but it was Today Yesterday).
Tuesday, January 6, 2015
Thursday, January 1, 2015
VAX/VMS was a wrong turn
VAX/VMS was a big part of my life in the 1980's. From 2003 to the end of 2008 I had a VT125 terminal on my office desk, and it was connected to a VAX in our machine room, which had many machines, including at least one 8600 if not an 8800 as well. The room after 2006 was a showpiece, visible through glass windows, with large climate control machines also.
The earlier room was also visible, but before the 8600 the mere 11/780's tied to many VT125's were dog slow, taking up to an hour just to log on. And that's where I became familiar with one VMS misfeature. The fact that while processing was being done, characters typed at your terminal were gleefully ignored. Basically, you'd watch, and if you didn't see your characters appearing after awhile, you'd type your command again. I recall one co-programmer (who later became a divinity student) intoned:
Having been thusly trained, when we moved on from 11/780's to 8600's the problem of having the terminal program swapped out while you were typing occurred less, but never totally went away. I recall forgetting about this feature, until a Unix programmer pointed it out to me again, and before long I was working on a Sun Unix system also and VAX/VMS and it's foibles were gradually forgotten.
DEC really earned their reputation on, at the time, leading machines, like the PDP-10 in 1966, the breakthrough timesharing machine, used by Compuserve, Stanford, MIT, and many others. This was not a perfect machine either, but in those days DEC was more honest, less driving by intense marketing. So their solution to imperfection was to provide the source code for the entire OS. This was a breakthrough for places like MIT, the PDP-10 became the delight of all "hackers" (as leading computer scientist students were often called then, before hacking became synonymous with cracking), since you could hack the very OS the great timesharing system was based on, and do virtually anything to it, though typical modifications were bypassing the serial input stuff, IIRC. Interesting. Meanwhile, for IBM machines, you slipped your PL/1 card deck through a slot and some time later it would come back with the first of many error reports. Big difference. Interactive multitasking was available early, but only to the college crowd, who later became the buyers of DEC machines like VAX, which simply had to be good, the best thing ever, etc. Except that it was way oversold, too slow for the timesharing it was sold for, and VMS was terrible.
I had a parallel experience during the PDP-11 "hacker" era. I was a enthusiastic PDP-10 user for all of my college years from 73 to 78. The PDP-10 user of the system for the Claremont Colleges, renowned as one of the last to keep going into the early 2000's. Though we only had a few dozen terminals on campus, it wasn't THAT hard to get on one, and it was just heaven, being able to solve difficult problems in Basic, or run cool things like Eliza. In the summer after my freshman year I did a programming job using APL through a Tektronics terminal (which did lots of funny characters) connected through the Dec 10, as we called it. The Dec 10 was so much more useful and fun than the IBM 360, we actually had in our math building at Pomona College. The Dec 10 was up campus in the Claremont University Center. Later I ran my psychology experiments through SPSS on the Dec 10.
But meanwhile, I had also come to know and love the PDP-11 I used for several years in the Psychology department. Starting in 1976, I programmed in Basic and Macro-11, thanks to a Macro-11 handbook we got. Using peeks and pokes, and finally Macro-11 programs, I controlled digital audio through a 10 bit multi channel audio interface, and graphical displays on a CRT.
RT-11 was wonderfully crisp, and the system was marvelously simple and open. For controlling laboratory equipment, nothing could be better.
I sadly missed out on the dawn of Unix, which also happened on PDP-11's, starting in 1970. Unix had originally been programmed on a PDP-7, something I've never seen, but the first C version was programmed on a PDP-11 in 1970, and PDP-11 became the standard through the key developments of Research System 6 and 7 and BSD. I only got to use Unix starting in 1989, after gratefully leaving behind VMS.
Last PDP-11 I programmed was an 11/60 running RSX-11m, which was not perfect but helpfully open and fixable.
VAX, sadly, was a step backwards from all these fine machines. Rather than a leader, it quickly became a laggard in speed. It was sold to do far more than it could. A single 11/780 might be a nice personal machine, or maybe two people. I don't believe DEC was as open with the OS as they previously had been, though the 780 was a departmental machine, maintained by specialists, who at least kept it working. Unix might have been available, but we were supporting our product under VMS, which was considered something like industry standard, if not high end (Silicon Graphics was the high end).
DEC believed so much in the VAX they cancelled their Jupiter project to create the next generation PDP-10. This made people using famous systems at Stanford very angry, and may have been the impetus to creating SUN (Stanford University Network) computers.
It was said that the successor to PDP-11/70 was killed because it was faster than the VAX. The same may have been true for the Jupiter. The VAX machines were slow and especially unresponsive especially with the 11/780 and 11/730. The improvement in the 11/785 was minimal. It took many years before the speed was substantially increased, when the 8600 came out. Meanwhile DEC had disc clustering, machine clustering (buy more slow machines!), but each machine could simply not work for many users running CAD programs.
I think VAX was a bad turn, not just for DEC. All systems have become either more closed or more complicated, to the point where systems like the wonderful PDP-11 don't make much sense anymore, but are very much missed.
One of the bigger parts of that mistake was VMS. DEC ultimately realized the error of their ways and made Ultrix their standard. As that was happening, I moved on to SUN computers, which had used BSD Unix from the beginning. Until the late 2000's, SUN occupied a position similar to that which DEC had had, in leading universities. With Oracle starting to actually charge for academic Solaris licenses, and the decline of the cost competitiveness of Sun desktop hardware, Gnu/Linux is gaining faster than ever to solidify it's position as the leading research OS. Mac OS X is unix based, and much unix code has made it into Microsoft Windows as well.
The earlier room was also visible, but before the 8600 the mere 11/780's tied to many VT125's were dog slow, taking up to an hour just to log on. And that's where I became familiar with one VMS misfeature. The fact that while processing was being done, characters typed at your terminal were gleefully ignored. Basically, you'd watch, and if you didn't see your characters appearing after awhile, you'd type your command again. I recall one co-programmer (who later became a divinity student) intoned:
The VAX, in it's infinite wisdom, sometimes chooses to ignore your command.Sometimes, just to keep you on your toes, the ignored character was precisely the carriage return (and it was CR, IIRC, not NL as on Unix machines) which caused the command to be executed. So you couldn't tell if it was just sitting there, after you entered the command, and slowly grinding away, or whether it has summarily chosen to ignore your CR.
Having been thusly trained, when we moved on from 11/780's to 8600's the problem of having the terminal program swapped out while you were typing occurred less, but never totally went away. I recall forgetting about this feature, until a Unix programmer pointed it out to me again, and before long I was working on a Sun Unix system also and VAX/VMS and it's foibles were gradually forgotten.
DEC really earned their reputation on, at the time, leading machines, like the PDP-10 in 1966, the breakthrough timesharing machine, used by Compuserve, Stanford, MIT, and many others. This was not a perfect machine either, but in those days DEC was more honest, less driving by intense marketing. So their solution to imperfection was to provide the source code for the entire OS. This was a breakthrough for places like MIT, the PDP-10 became the delight of all "hackers" (as leading computer scientist students were often called then, before hacking became synonymous with cracking), since you could hack the very OS the great timesharing system was based on, and do virtually anything to it, though typical modifications were bypassing the serial input stuff, IIRC. Interesting. Meanwhile, for IBM machines, you slipped your PL/1 card deck through a slot and some time later it would come back with the first of many error reports. Big difference. Interactive multitasking was available early, but only to the college crowd, who later became the buyers of DEC machines like VAX, which simply had to be good, the best thing ever, etc. Except that it was way oversold, too slow for the timesharing it was sold for, and VMS was terrible.
I had a parallel experience during the PDP-11 "hacker" era. I was a enthusiastic PDP-10 user for all of my college years from 73 to 78. The PDP-10 user of the system for the Claremont Colleges, renowned as one of the last to keep going into the early 2000's. Though we only had a few dozen terminals on campus, it wasn't THAT hard to get on one, and it was just heaven, being able to solve difficult problems in Basic, or run cool things like Eliza. In the summer after my freshman year I did a programming job using APL through a Tektronics terminal (which did lots of funny characters) connected through the Dec 10, as we called it. The Dec 10 was so much more useful and fun than the IBM 360, we actually had in our math building at Pomona College. The Dec 10 was up campus in the Claremont University Center. Later I ran my psychology experiments through SPSS on the Dec 10.
But meanwhile, I had also come to know and love the PDP-11 I used for several years in the Psychology department. Starting in 1976, I programmed in Basic and Macro-11, thanks to a Macro-11 handbook we got. Using peeks and pokes, and finally Macro-11 programs, I controlled digital audio through a 10 bit multi channel audio interface, and graphical displays on a CRT.
RT-11 was wonderfully crisp, and the system was marvelously simple and open. For controlling laboratory equipment, nothing could be better.
I sadly missed out on the dawn of Unix, which also happened on PDP-11's, starting in 1970. Unix had originally been programmed on a PDP-7, something I've never seen, but the first C version was programmed on a PDP-11 in 1970, and PDP-11 became the standard through the key developments of Research System 6 and 7 and BSD. I only got to use Unix starting in 1989, after gratefully leaving behind VMS.
Last PDP-11 I programmed was an 11/60 running RSX-11m, which was not perfect but helpfully open and fixable.
VAX, sadly, was a step backwards from all these fine machines. Rather than a leader, it quickly became a laggard in speed. It was sold to do far more than it could. A single 11/780 might be a nice personal machine, or maybe two people. I don't believe DEC was as open with the OS as they previously had been, though the 780 was a departmental machine, maintained by specialists, who at least kept it working. Unix might have been available, but we were supporting our product under VMS, which was considered something like industry standard, if not high end (Silicon Graphics was the high end).
DEC believed so much in the VAX they cancelled their Jupiter project to create the next generation PDP-10. This made people using famous systems at Stanford very angry, and may have been the impetus to creating SUN (Stanford University Network) computers.
It was said that the successor to PDP-11/70 was killed because it was faster than the VAX. The same may have been true for the Jupiter. The VAX machines were slow and especially unresponsive especially with the 11/780 and 11/730. The improvement in the 11/785 was minimal. It took many years before the speed was substantially increased, when the 8600 came out. Meanwhile DEC had disc clustering, machine clustering (buy more slow machines!), but each machine could simply not work for many users running CAD programs.
I think VAX was a bad turn, not just for DEC. All systems have become either more closed or more complicated, to the point where systems like the wonderful PDP-11 don't make much sense anymore, but are very much missed.
One of the bigger parts of that mistake was VMS. DEC ultimately realized the error of their ways and made Ultrix their standard. As that was happening, I moved on to SUN computers, which had used BSD Unix from the beginning. Until the late 2000's, SUN occupied a position similar to that which DEC had had, in leading universities. With Oracle starting to actually charge for academic Solaris licenses, and the decline of the cost competitiveness of Sun desktop hardware, Gnu/Linux is gaining faster than ever to solidify it's position as the leading research OS. Mac OS X is unix based, and much unix code has made it into Microsoft Windows as well.
Wednesday, December 31, 2014
Mathematicians and their insane symbology
First of all, mathematicians never use descriptive names in their formulae. They use letters, often starting with greek letters, then when they have exhausted all the greek letters they pull out other letters from other languages.
Most of us don't have keyboards with all those letters, or that can even reproduce them easily. Let alone having learned them or not. Sometimes the letters are barely distinguishable from other letters in the same formula, from another language perhaps. And the names don't have meaning to the uninitiated, though that part may be the least bad, as over time one becomes so used to the greek letter, or whatever, that the technical name is barely used anymore or maybe even never well learned. So it is with sigma, rho, and a few others, at least for me.
Then the letters can have an endless variety of qualifying marks, little hats, tick marks at varying angles, circles, crosses, bolded crosses, dots, underlines. Then you can have subscripting, superscripting, repeated superscripting.
This of course provides a way to make the most complicated formulae compact. It enables symbolic operations, and writing things down with variations to get at some end (though mathematicians and many math books forgo many if not all intermediate steps anyway).
So it may indeed be good, in most ways, for the professional, for it to be like this.
It's generally not good for explaining things to people who are not professional mathematicians.
Most of us don't have keyboards with all those letters, or that can even reproduce them easily. Let alone having learned them or not. Sometimes the letters are barely distinguishable from other letters in the same formula, from another language perhaps. And the names don't have meaning to the uninitiated, though that part may be the least bad, as over time one becomes so used to the greek letter, or whatever, that the technical name is barely used anymore or maybe even never well learned. So it is with sigma, rho, and a few others, at least for me.
Then the letters can have an endless variety of qualifying marks, little hats, tick marks at varying angles, circles, crosses, bolded crosses, dots, underlines. Then you can have subscripting, superscripting, repeated superscripting.
This of course provides a way to make the most complicated formulae compact. It enables symbolic operations, and writing things down with variations to get at some end (though mathematicians and many math books forgo many if not all intermediate steps anyway).
So it may indeed be good, in most ways, for the professional, for it to be like this.
It's generally not good for explaining things to people who are not professional mathematicians.
Monday, December 29, 2014
The Missing Today
I think I see what's happening. It's 2:30 am Monday and the low expected in the next 12 hours is 34F, around 6am, according to the projected temperature graph (and yesterday's forecast of "today", as of 11pm).
But you wouldn't know that looking at the "Today" projections at Weather Underground. It predicts a low of 41F. Which is what the graph says it will be at around 6am on Tuesday.
OK, so that fits the intuitive notion of "today" (though I don't see that spelled out) meaning that the low is in the night following the day. But what about the projected low in the next 3 hours, that was "Today" until 11 am. Then at midnight, Today became Yesterday, but Yesterday seems to show a low of 36, which may have been the low on Saturday morning at 6am.
So in my view it's OK to have an intuitive notion of day, and in fact I'd start each day at 6am. But if you start doing that sort of thing, you ought to be consistent or else "Today" disappears within a certain range of hours, as it has right now.
But you wouldn't know that looking at the "Today" projections at Weather Underground. It predicts a low of 41F. Which is what the graph says it will be at around 6am on Tuesday.
OK, so that fits the intuitive notion of "today" (though I don't see that spelled out) meaning that the low is in the night following the day. But what about the projected low in the next 3 hours, that was "Today" until 11 am. Then at midnight, Today became Yesterday, but Yesterday seems to show a low of 36, which may have been the low on Saturday morning at 6am.
So in my view it's OK to have an intuitive notion of day, and in fact I'd start each day at 6am. But if you start doing that sort of thing, you ought to be consistent or else "Today" disappears within a certain range of hours, as it has right now.
Tuesday, December 16, 2014
Tale of Two Spell Checkers
The spell checker in the Mac Mail tool that I use at work is constantly changing the technical terms I use (in my work) into other words, fundamentally destroying what I am attempting to communicate. It does this automatically requiring an explicit click to defeat (requiring a reach for the mouse--a big nuisance when you have two hands on keyboard) and I sometimes overlook this. No matter how many times I use a particular term unknown to it (which could be a private or common acronym, a computer language command, or an esoteric concept word) the spell checker never learns. Even when the technical terms I use are in all letters, and may actually be valid if obscure english words, and I use them frequently, they aren't remembered. Right now, typing into Blogger from another Mac, a similar if not identical checker seems to be working, so maybe it's Mac specific rather than tool specific (now I am using Safari, into Blogger). I recall the spell checker in my iPhone working similarly, but it wasn't as much a nuisance with that, since I am rarely giving technical support via phone, and even if I did it would be very short. Actually, on the phone, where misspelling was far too easy compared with keyboard at work, an aggressive spell checker with limited vocabulary may have been ideal.
The spell checker in my Android based Samsung Galaxy S4 message tool seems much different, and very inconveniently so in a different way. It seems to have pre-memorized all the typical mistakes made on an Android keyboard, and added them to an official word list. Especially, on the tiny keyboard, it's all too easy to type a number when you mean a letter. So for example I am trying to enter "work" but enter "2ork" instead. Then it doesn't show "work" as a correction choice (factoring in the proximity of numbers to the top row of letters). It shows me 2ork (what I just entered wrongly!) 23rd, and 20th. On the longer list of possibilities, not even one starts with a w, the letter right below 2 on the keyboard. I've often entered digits invalidly in words and had the message tool checker give me all similar looking nonsense choices with digits in the middle of words. And the same is true with many all-letter misspellings as well. The choices I am offered are all horribly misspelled words--most often not sounding like any acronym I've ever heard. And when one of my horribly misspelled words goes through (as happens all too often) I believe it will be around to haunt me even more in the future as a spell checker preferred choice.
And as it happens, the Galaxy S4 mail tool keyboard (portrait mode) has the spell corrected words just above the numbers…so I might just type another number when I intended to choose a spelling correction to a word with one number. And then I will get only nonsense choices, with two or more numbers in them, or perhaps only just what I entered. How useful is that?
As with so many computer "conveniences" one wonders how much the designers actually used these designs before foisting them on the public. Or how many kinds of other people they got feedback from in refining the design. It seems often that flaws in technical products should have been noticed and corrected in the very first prototype. Or not made in the first place by a thoughtful designer.
But then if designs are created only from the top down, from directives like "make a spell checker which learns new words and presents choices", without much further thought or testing on the part of the designers and others working from that directive, we might get these kinds of deeply flawed products.
The spell checker in my Android based Samsung Galaxy S4 message tool seems much different, and very inconveniently so in a different way. It seems to have pre-memorized all the typical mistakes made on an Android keyboard, and added them to an official word list. Especially, on the tiny keyboard, it's all too easy to type a number when you mean a letter. So for example I am trying to enter "work" but enter "2ork" instead. Then it doesn't show "work" as a correction choice (factoring in the proximity of numbers to the top row of letters). It shows me 2ork (what I just entered wrongly!) 23rd, and 20th. On the longer list of possibilities, not even one starts with a w, the letter right below 2 on the keyboard. I've often entered digits invalidly in words and had the message tool checker give me all similar looking nonsense choices with digits in the middle of words. And the same is true with many all-letter misspellings as well. The choices I am offered are all horribly misspelled words--most often not sounding like any acronym I've ever heard. And when one of my horribly misspelled words goes through (as happens all too often) I believe it will be around to haunt me even more in the future as a spell checker preferred choice.
And as it happens, the Galaxy S4 mail tool keyboard (portrait mode) has the spell corrected words just above the numbers…so I might just type another number when I intended to choose a spelling correction to a word with one number. And then I will get only nonsense choices, with two or more numbers in them, or perhaps only just what I entered. How useful is that?
As with so many computer "conveniences" one wonders how much the designers actually used these designs before foisting them on the public. Or how many kinds of other people they got feedback from in refining the design. It seems often that flaws in technical products should have been noticed and corrected in the very first prototype. Or not made in the first place by a thoughtful designer.
But then if designs are created only from the top down, from directives like "make a spell checker which learns new words and presents choices", without much further thought or testing on the part of the designers and others working from that directive, we might get these kinds of deeply flawed products.
Saturday, November 8, 2014
Friday, September 12, 2014
Bread having seeds should have warning label
I always go for the most whole wheat of breads as they are the most healthy. But in the healthy spectrum of bread, something that's all too popularly sold in my opinion is bread with seeds. Somehow seeds are supposed to be associated with something healthy. But I have never heard that seeds are an essential part of diet, they may not be well digested, and they get stuck in teeth (that happens to me a lot).
Every once in awhile I find myself buying bread with seeds because it's not necessarily clear from the packaging which bread has seeds and which doesn't. Then, when I do, about 4 days later I have serious gum pain which lasts for days even if I remember to cut out the bread with seeds.
Since they don't much advertise which breads have seeds and which don't, or even have any markings on the packaging, it seems like the plan is to dump seeds on customers, as though they were some waste product which is most profitably discharged by mixing it with the product sold.
My latest mistake was "Oroweat 12 Grains" whose seeds made my gums hurt in 3 days. I don't see that it warns about seeds anywhere on the package. My usual bread is Oroweat 100% Whole Wheat. Oroweat used to make their top 100% Whole Wheat bread with Stone Ground Whole Wheat, but this has disappeared in the list of ingredients. Still, the bread has texture more like stone ground than most. A doctor in 1973 advised me to always find "the real whole wheat, stone ground whole wheat." And by now that is not easy anymore.
Every once in awhile I find myself buying bread with seeds because it's not necessarily clear from the packaging which bread has seeds and which doesn't. Then, when I do, about 4 days later I have serious gum pain which lasts for days even if I remember to cut out the bread with seeds.
Since they don't much advertise which breads have seeds and which don't, or even have any markings on the packaging, it seems like the plan is to dump seeds on customers, as though they were some waste product which is most profitably discharged by mixing it with the product sold.
My latest mistake was "Oroweat 12 Grains" whose seeds made my gums hurt in 3 days. I don't see that it warns about seeds anywhere on the package. My usual bread is Oroweat 100% Whole Wheat. Oroweat used to make their top 100% Whole Wheat bread with Stone Ground Whole Wheat, but this has disappeared in the list of ingredients. Still, the bread has texture more like stone ground than most. A doctor in 1973 advised me to always find "the real whole wheat, stone ground whole wheat." And by now that is not easy anymore.
Subscribe to:
Posts (Atom)