Re: [gutvol-d] Addition to PG History PT1
*sigh* The only thing your chart proves is that applying Moore's Law to PG production is a waste of time. 1991 is NOT when production started. It is not when PG started. It is just the point in time you chose as the starting point to attach Moore's Law. What people are saying is this: MOORE'S LAW DOES NOT APPLY TO PG PRODUCTION. WE WOULD BE BETTER OFF JUST ANNOUNCING WE HAVE "X" NUMBER OF TEXTS INSTEAD OF COMPARING IT TO A "LAW" THAT ISN'T EVEN MEANT TO BE USED THIS WAY. Admit it, Michael, 1991 is just an arbritrary date. Just because "regular production" started in 1991 ... bull. We had regular, once a year, production before that. Josh PS BTW, if you are going to say no one "read the message above," you might want to actually quote what we supposedly didn't read. As it is, I'm assuming we all read the same thing as you, but had a slightly different level of reading comprehension, since we all seemed to get something completely different from it than you did. ----- Original Message ----- From: "Michael Hart" <hart@pglaf.org> To: "The gutvol-d Mailing List" <gutvol-d@lists.pglaf.org> Subject: [gutvol-d] Addition to PG History PT1 Date: Tue, 25 Jan 2005 07:34:58 -0800 (PST)
Well, it would appear those in question did not read the message above, at least they did not reply to any of the points made, so I am duty bound to explain in a bit more detail, as to how there are no other choices other than Project Gutenberg's first effort at a regular production schedule in 1991 as a starting point.
Here are various starting years and results for Moore's Law used to project the growth of Project Gutenberg.
START TOTAL START TOTAL START TOTAL START TOTAL YEAR NUMBER YEAR NUMBER YEAR NUMBER YEAR NUMBER ACTUAL 1971 BASE 1971 1979 BASE 1979 1990 BASE 1990 1993 BASE 1993 NUMBER/YR
1971 1 1 1971 1974 4 4 1974 1977 16 7 1977 1980 64 1979 9 9 1980 1983 256 1982 36 9 1983 1986 1024 1985 144 10 1986 1989 4096 1988 576 1990 10 10 1989 1992 16384 1991 2304 1993 40 1993 100 42 1992 1995 65536 1994 9216 1996 160 1996 400 365 1995 1998 262144 1997 36864 1999 640 1999 1600 1550 1998 2001 1480576 2000 147456 2002 2560 2002 25600 4260 2001 2004 4194304 2003 589824 2005 10240 2005 102400 14944 2004 2007 16777216 2006 2359296 2008 40960 2008 409600 ????? 2007
As you can plainly see, starting Moore's Law at any other date than ~1991 is inappropriate not only because the numbers make it obvious, but also because 1991 was the first year of regular production growth for Project Gutenberg.
_______________________________________________ gutvol-d mailing list gutvol-d@lists.pglaf.org http://lists.pglaf.org/listinfo.cgi/gutvol-d
I would agree that choosing 1991 as a starting point for a Moore's Law relation on Project Gutenberg production is arbitrary. At the same time, however, let's think about using this same logic for the semiconductor industry. Rather than choosing some arbitrary starting point for our measurements, let's go straight back to the beginning. The first germanium transistor was invented in 1947, and it had a minimum feature size of 0.002 inches, or about 5.08x10^-5 meters. According to Moore's law, this feature size should reduce by half roughly every 18 months, correct? By that prediction, the feature size should have since been reduced by half 38.667 times. That would give us a minimum feature size in today's transistors of 1.16x10^-16 meters, or 0.000001 Angstroms. Of course, the diameter of an atom is on the scale of Angstoms, meaning that under Moore's Law we should now be fitting nearly one million transistors within a single atom...needless to say we are nowhere close. Therefore, the semiconductor industry is only keeping up with Moore's Law in as much as Project Gutenberg is. I didn't take the time to compare other features of the first semiconductor transistor (or the first integrated circuit), as I had a bit more trouble locating other data, but I suspect the result would be the same. The point, of course, is that Moore's Law serves to indicate an exponential growth rate over whatever time frame you choose to apply it. Regardless of where it starts, it gives a pretty good idea of this. Do I think that it is particularly applicable to Project Gutenberg production? No, not really. But it is a term that the public is familiar with, and thus Michael finds that it is a useful term to use in advertising just how much work PG has been doing over the past decade (and a half). Can we really think in terms of "keeping up with Moore's Law?" Probably not, it doesn't make much sense, though the semiconductor industry does it every day, and going back to the very beginning, they are much farther away from Moore's Law than we are. - Scott Schmucker Joshua Hutchinson wrote:
*sigh*
The only thing your chart proves is that applying Moore's Law to PG production is a waste of time.
1991 is NOT when production started. It is not when PG started. It is just the point in time you chose as the starting point to attach Moore's Law.
What people are saying is this:
MOORE'S LAW DOES NOT APPLY TO PG PRODUCTION. WE WOULD BE BETTER OFF JUST ANNOUNCING WE HAVE "X" NUMBER OF TEXTS INSTEAD OF COMPARING IT TO A "LAW" THAT ISN'T EVEN MEANT TO BE USED THIS WAY.
Admit it, Michael, 1991 is just an arbritrary date. Just because "regular production" started in 1991 ... bull. We had regular, once a year, production before that.
Josh
PS BTW, if you are going to say no one "read the message above," you might want to actually quote what we supposedly didn't read. As it is, I'm assuming we all read the same thing as you, but had a slightly different level of reading comprehension, since we all seemed to get something completely different from it than you did.
----- Original Message ----- From: "Michael Hart" <hart@pglaf.org> To: "The gutvol-d Mailing List" <gutvol-d@lists.pglaf.org> Subject: [gutvol-d] Addition to PG History PT1 Date: Tue, 25 Jan 2005 07:34:58 -0800 (PST)
Well, it would appear those in question did not read the message above, at least they did not reply to any of the points made, so I am duty bound to explain in a bit more detail, as to how there are no other choices other than Project Gutenberg's first effort at a regular production schedule in 1991 as a starting point.
Here are various starting years and results for Moore's Law used to project the growth of Project Gutenberg.
START TOTAL START TOTAL START TOTAL START TOTAL YEAR NUMBER YEAR NUMBER YEAR NUMBER YEAR NUMBER ACTUAL 1971 BASE 1971 1979 BASE 1979 1990 BASE 1990 1993 BASE 1993 NUMBER/YR
1971 1 1 1971 1974 4 4 1974 1977 16 7 1977 1980 64 1979 9 9 1980 1983 256 1982 36 9 1983 1986 1024 1985 144 10 1986 1989 4096 1988 576 1990 10 10 1989 1992 16384 1991 2304 1993 40 1993 100 42 1992 1995 65536 1994 9216 1996 160 1996 400 365 1995 1998 262144 1997 36864 1999 640 1999 1600 1550 1998 2001 1480576 2000 147456 2002 2560 2002 25600 4260 2001 2004 4194304 2003 589824 2005 10240 2005 102400 14944 2004 2007 16777216 2006 2359296 2008 40960 2008 409600 ????? 2007
As you can plainly see, starting Moore's Law at any other date than ~1991 is inappropriate not only because the numbers make it obvious, but also because 1991 was the first year of regular production growth for Project Gutenberg.
_______________________________________________ gutvol-d mailing list gutvol-d@lists.pglaf.org http://lists.pglaf.org/listinfo.cgi/gutvol-d
_______________________________________________ gutvol-d mailing list gutvol-d@lists.pglaf.org http://lists.pglaf.org/listinfo.cgi/gutvol-d
Scott Schmucker wrote:
The point, of course, is that Moore's Law serves to indicate an exponential growth rate over whatever time frame you choose to apply it. Regardless of where it starts, it gives a pretty good idea of this. Do I think that it is particularly applicable to Project Gutenberg production? No, not really. But it is a term that the public is familiar with, and thus Michael finds that it is a useful term to use in advertising just how much work PG has been doing over the past decade (and a half). Can we really think in terms of "keeping up with Moore's Law?" Probably not, it doesn't make much sense, though the semiconductor industry does it every day, and going back to the very beginning, they are much farther away from Moore's Law than we are.
In the few times I tried to apply Moore's Law to something other than CPU advancement, I've used the phrase "Moore's Law-like." However, I think we should use the term "exponential" or "geometric" (whatever best applies) to describe the growth rate of PG texts. Jon
On Wed, 26 Jan 2005, Scott Schmucker wrote:
I would agree that choosing 1991 as a starting point for a Moore's Law relation on Project Gutenberg production is arbitrary.
Arbitary refers to the choice of something without any particular reason for the decision and there are two very non-arbitary reasons for choosing 1991 that have been mentioned quite thoroughly without being refuted by any better model. Once other models were shown to have been chosen to reflect points from which large deviations from both our goals and from reality were predicted [i.e. proven to be false] the argument changed from picking a different year to picking no year at all. [snipped semiconductor history]
The point, of course, is that Moore's Law serves to indicate an exponential growth rate over whatever time frame you choose to apply it. Regardless of where it starts, it gives a pretty good idea of this. Do I think that it is particularly applicable to Project Gutenberg production? No, not really. But it is a term that the public is familiar with, and thus Michael finds that it is a useful term to use in advertising just how much work PG has been doing over the past decade (and a half). Can we really think in terms of "keeping up with Moore's Law?" Probably not, it doesn't make much sense, though the semiconductor industry does it every day, and going back to the very beginning, they are much farther away from Moore's Law than we are.
So. . .when it comes down to it, Moore's Law applies better in our case than in the case of semiconductors. However, I should again point out that here it is just being used as a goal and a reference point, and has been for 15 years now. It would appeare that as long as we were always well ahead of Moore's Law for each of our projections, that no one was going to complain a lot, but now that a year has been reported that fell short, even of the previous year, much less of the previous 18 months, then the noise level increases. If there were truly an aversion to using Moore's Law for these purposes, this aversion would likely have been brought up quite often from the 1991 to the present. Of course, you are welcome to apply any techniques to your own history of Project Gutenberg, and run them up our flagpole for testing. Michael
Michael Hart wrote:
I would agree that choosing 1991 as a starting point for a Moore's Law relation on Project Gutenberg production is arbitrary.
Arbitary refers to the choice of something without any particular reason for the decision and there are two very non-arbitary reasons for choosing 1991 that have been mentioned quite thoroughly without being refuted by any better model.
The one reason being that any other year was proven not to work. The other reason I forgot. I'm against "Laws" that work only on fridays with full moon. -- Marcello Perathoner webmaster@gutenberg.org
#caps lock ON OK. THAT'S ENOUGH ALREADY. WE GET THE POINT. LET'S GIVE THIS TOPIC A REST. NO MORE DISCUSSION OF MOORE'S LAW AND PG. PLEASE.................. #caps lock OFF
Joshua Hutchinson wrote:
*sigh*
The only thing your chart proves is that applying Moore's Law to PG production is a waste of time.
From the apparent perspective of your comments, this entire discussion would be considered a waste of time, so I'm wondering at the point of all you have said here.
1991 is NOT when production started.
Note the elimination of the word "regular" from "regular production." This sort of misquoting is hard on the reputation of the speaker.
It is not when PG started.
It is when PG first started a goal of regularly increasing production.
It is just the point in time you chose as the starting point to attach Moore's Law.
It is the starting point of attaching ANY kind of predictive PG goals. Moore's Law just happened to be handy, and to fit with what I though Project Gutenberg could do in the future. All in all, it's been remarkable how well starting Moore's Law from 1991 has worked. I keep asking for a better model of prediction, and having to spend my time proving how inelegantly the suggested years work out when the equations are actually moved from their elegant non-numerical form into real numbers that are obviously out of the realm of reality.
What people are saying is this:
MOORE'S LAW DOES NOT APPLY TO PG PRODUCTION.
Then why does it fit with reality so much better than anything else?
WE WOULD BE BETTER OFF JUST ANNOUNCING WE HAVE "X" NUMBER OF TEXTS INSTEAD OF COMPARING IT TO A "LAW" THAT ISN'T EVEN MEANT TO BE USED THIS WAY.
There have been many technical reports filed on who should be able to use Moore's Law, in which manners it should be used, what fields it should and should not be allowed to be used in. . . but the reality of the situation is that Moore's Law has been out of the realm of the technical experts for most of its history, and you can't just stuff it back in the bottle.
Admit it, Michael, 1991 is just an arbritrary date.
Again I refer you to the opening definitions of arbitrary.
Just because "regular production" started in 1991 ... bull.
Ah, now you use the accurate quotation, and have lost your language skills.
We had regular, once a year, production before that.
If we take your statement of half truth at face value, by your count 1971 to 1990 would then yield 20 entries. Again, focusing only oh a half truth, leaving out the rest, digs you further into a hole. 9 years of "regular, once a year, production before that" followed by 11 years in which only one number was added, includes both halves of your truth, and also indicates why 1991 is the first year from which to make projections. At least projections that are not linear, at best. * However, the real point is that all this information has been presented before, then reflected back in distortion, which I have taken time to very politely refute, time and time again, without resorting to attacking the person and not the ideas presented, and suggesting how the argument might be better made to successfully make your points. Obvious attempts have been made to make this personal, which have been ignored, and the most obvious attempt is to remove any measuring stick for our progress. As with the failed suggested models of starting 1971 and 1993 as the best baselines for Moore's Law, the idea of removing any such yardstick at all is merely a ploy to remove any objective measurement standard. Whether we meet such a standard, exceed it, or even fall short of it, it is always handy to have such a standard so we know something about where are are, where we came from, and where we are going. Michael
Before December 1990, PG's collection size did not grow exponentially. Between December 1990 and August 1997, the collection size grew exponentially with a 12-month doubling time. Between August 1997 and August 2000, the collection size grew exponentially with a 27-month doubling time. Between August 2000 and December 2004, the collection size grew exponentialy with a 20-month doubling time. On average, between December 1990 and December 2004, the collection size grew exponentially with a 16-month doubling time. 1991 is a fine year to begin exponential fitting, but our present rate of exponential growth is somewhat less than it has been historically. -- RS
Michael Hart wrote:
I keep asking for a better model of prediction,
If you want a 'doubling every 18 months' curve, then using a reference point of 100 in 1993 actually provides a much better fit to the (post-1990) data than does 10 in 1990.
... the failed suggested models of starting 1971 and 1993 as the best baselines for Moore's Law, ...
Re the "failure" of the 1993 model, in yesterday's posting: http://lists.pglaf.org/private.cgi/gutvol-d/2005-January/001419.html you gave this data: (omitting the columns that use 1971 and 1979 as reference points) START TOTAL START TOTAL YEAR NUMBER YEAR NUMBER ACTUAL 1990 BASE 1990 1993 BASE 1993 NUMBER/YR 1 1971 4 1974 7 1977 9 1980 9 1983 10 1986 1990 10 10 1989 1993 40 1993 100 42 1992 1996 160 1996 400 365 1995 1999 640 1999 1600 1550 1998 2002 2560 2002 25600 4260 2001 2005 10240 2005 102400 14944 2004 2008 40960 2008 409600 ????? 2007 However, there's a mistake in the "BASE 1993" column: the numbers after 1600 should be 6400, 25600, and 102400. Moreover, it's easier to make the comparison if we use the same years for the actual numbers as we do for the predicted numbers. With these two changes, I think the table should be something like: ref point = ref point = year '10 in 1990' '100 in 1993' actual 1990 10 25 10 1993 40 100 100 1996 160 400 750 1999 640 1600 2000 2002 2560 6400 6500 2005 10240 25600 20000? 2008 40960 102400 ????? (It should also be clarified that the number for a given year is that of Dec 10th for that year.) As you can see, the actual numbers (after 1990) are closer to the "1993" curve than the "1990" curve. -Michael Dyck ps: In your table, there's also a typo in the "BASE 1971" column: 1480576 should be 1048576.
Fixed the typos, thanks so much!!! I think I must have gotten a double click when running the "Base 1993". . .thanks for catching that. . .been working too hard lately, or [hopefully] I would see those errors myself. Will work on other year models as well. More thanks!!! Michael On Wed, 26 Jan 2005, Michael Dyck wrote:
Michael Hart wrote:
I keep asking for a better model of prediction,
If you want a 'doubling every 18 months' curve, then using a reference point of 100 in 1993 actually provides a much better fit to the (post-1990) data than does 10 in 1990.
... the failed suggested models of starting 1971 and 1993 as the best baselines for Moore's Law, ...
Re the "failure" of the 1993 model, in yesterday's posting: http://lists.pglaf.org/private.cgi/gutvol-d/2005-January/001419.html you gave this data: (omitting the columns that use 1971 and 1979 as reference points)
START TOTAL START TOTAL YEAR NUMBER YEAR NUMBER ACTUAL 1990 BASE 1990 1993 BASE 1993 NUMBER/YR 1 1971 4 1974 7 1977 9 1980 9 1983 10 1986 1990 10 10 1989 1993 40 1993 100 42 1992 1996 160 1996 400 365 1995 1999 640 1999 1600 1550 1998 2002 2560 2002 25600 4260 2001 2005 10240 2005 102400 14944 2004 2008 40960 2008 409600 ????? 2007
However, there's a mistake in the "BASE 1993" column: the numbers after 1600 should be 6400, 25600, and 102400. Moreover, it's easier to make the comparison if we use the same years for the actual numbers as we do for the predicted numbers. With these two changes, I think the table should be something like:
ref point = ref point = year '10 in 1990' '100 in 1993' actual
1990 10 25 10 1993 40 100 100 1996 160 400 750 1999 640 1600 2000 2002 2560 6400 6500 2005 10240 25600 20000? 2008 40960 102400 ?????
(It should also be clarified that the number for a given year is that of Dec 10th for that year.)
As you can see, the actual numbers (after 1990) are closer to the "1993" curve than the "1990" curve.
-Michael Dyck
ps: In your table, there's also a typo in the "BASE 1971" column: 1480576 should be 1048576.
_______________________________________________ gutvol-d mailing list gutvol-d@lists.pglaf.org http://lists.pglaf.org/listinfo.cgi/gutvol-d
participants (8)
-
John Hagerson -
Jon Noring -
Joshua Hutchinson -
Marcello Perathoner -
Michael Dyck -
Michael Hart -
Robert Shimmin -
Scott Schmucker