I strive for accuracy and clarity in my GammaFAQ, my ColorFAQ, in my book, and in my other publications. I'm no genius, but I'm a fairly smart guy, and I have a fairly decent reputation in the science of image coding. In addition to enjoying writing, I enjoy teaching; I teach many courses and seminars. For me, the reward of teaching is what I learn from it! As any attendee of any of my courses will tell you, I especially enjoy naïve questions.
As a result of publishing the FAQs, very frequently, at least several times a week, I communicate - sometimes privately, sometimes publicly - with someone who asks a question or makes a comment about some aspect of image coding or color reproduction. Often, the questions are already answered in my FAQs, my book, articles at my web site, or in papers that I have presented at various conferences. Sometimes a question involves a topic outside my area of expertise, and I direct the questioner to a more appropriate source of information. Once in a while these communications result in very interesting discussions, and sometimes improvements to the Gamma FAQ or Color FAQ result. On these occasions, both my correspondent I and learn from each other.
Once in a while, perhaps annually, I engage in communication - sometimes privately, sometimes publicly - with someone who questions my integrity or my objectivity. Twice, such communication has degenerated into a skirmish. On both occasions I tried to keep the skirmish private (at least until its resolution); on both occasions my correspondent "went public" in the Usenet newsgroups.
The first occasion was in mid 1994. David Bourgin had written and published on the Internet a "colorspace-faq." It was widely distributed, accepted as an Internet FAQ, and archived at the MIT/RTFM FAQ server. It was riddled with errors. I won't bore you with the details, but Bourgin was intransigent in ignoring requests from me and from 5 or 6 other experts to repair things. Eventually, Bourgin relinquished his FAQ document, and I replaced it with my Gamma FAQ and Color FAQ. My document is now the "official" Usenet colorspace-faq (as much as anything on Usenet can be "official"). Many copies of the obsolete Bourgin document still reside on servers around the world, and many of those obsolete files remain publicly-accessible.
The second occasion arose in mid February, 1998. Timo Autiokari posted in a public newsgroup - and sent privately - an open letter that started (sarcastically or not, I'm not sure),
Dear Mr. Poynton,
I find your explanation of the vision system versus the monitor gamma in the section 5 of your GammaFAQ to be a real pearl. ...
He then went on to ask several convoluted questions about the Gamma FAQ, and demonstrated some confusion about why nonlinear coding is important. He is obviously quite knowledgeable about many aspects of hardware and software.
Concerning image coding, he understands very well that intensity values in front of the camera must eventually produce comparable intensity values on the display; so, he understands that the end-to-end transfer function of an imaging system must be more or less linear - a straight-line function. However, he has apparently drawn the incorrect conclusion that the spacing of the code-points along that line has to be uniform. He seems not to comprehend the importance of Weber's law: He seems not to understand that along this line, the code points must be nonlinearly spaced to achieve good perceptual performance in an 8-bit system. Here's the situation:
I pointed out to Timo that the nonlinear coding is implemented in all video systems (analog, digital, and HDTV), Kodak's PhotoCD, the PNG file format, and the sRGB specification (from Silicon Graphics, Hewlett-Packard, and Microsoft).
In his posts, Timo has also demonstrated ignorance of the internals architecture and electronics of digital still cameras. He imagines that a camera advertised as having 8 bits must deliver 8 bits of linearly-coded intensity information. He does not understand that most - or perhaps even all! - of these cameras employ nonlinear coding. This issue was clarified in the newsgroups by a definitive posting from a gentleman from Hewlett-Packard Labs. I pointed out to Timo that nonlinear coding, more or less exactly as I describe it, is embedded in the new FlashPix format for digital still cameras, developed and adopted by Kodak, Hewlett-Packard, Microsoft, and others.
Timo remained unconvinced.
I also pointed out to Timo that Weber's law is not just applicable to image intensity: Audio intensity is expressed in units of decibels (a nonlinear measure), and telephone speech coding is nonlinear (according to A-law or -law).
Timo is not swayed.
Normally, this situation wouldn't bother me at all, even if played out in a public forum - I might even find it mildly amusing. What makes the situation rather special for me is that - somewhat like the Bourgin situation - Timo attributed, and perhaps continues to attribute, malicious intent. He published on 1997-02-25 (and withdrew, on 1997-02-26) a long web page that commenced (emphasis in original):
Gamma FAQ, ColorFAQ and GammaFQA by Mr. Poynton are fundamentally incorrect.
There is a fundamental error in Mr. Poynton's on-line documents that affects to most of his conclusions.
Due to the heavy terminology-play that predominate his documents, the error may seem to be insignificant or even non-existent. It is un-noticeable to many. Mr. Poynton's documents have mislead even some of the people at color laboratories of the major players in digital imaging business.
Well, I must be quite powerful and quite nefarious indeed, in order to mislead the most senior and most highly respected color researchers at Kodak, Hewlett-Packard, Silicon Graphics, and Microsoft. And they must be very gullible!
His Usenet news posting was clear enough, but in private e-mail sent on Mon, 23 Feb 1998 at 17:06:59 GMT, he was even more vehement, claiming that I had intentionally made the Gamma FAQ miselading.
The wonderful thing and the terrible thing about the net is that any information posted there - in a Usenet news article, or on a web page - is only as credible as its author. I think I'm credible, and I think Timo is incredible. I certainly have no reason to suspect or believe that Timo is a liar, a crook, or a scam artist. However, aside from the liars, crooks, and scam artists, every author considers himself to be credible, and on the net, everyone is an author! Anyone can publish! In a subject area in which yourself are not expert, how do you, the reader, determine who's credible? How do you identify who the true experts are? What do you do when two supposed experts disagree on something fundamental? (Additional information is available about Evaluating information on the web.)
Well of course you can rely upon reputation, and that's good starting point. But sometimes even highly respected people or sources make mistakes, so you must always keep an open mind. If you're a scientist, you must always be cognizant of the scientific method, and be on the lookout for experiments that prove or disprove the assertions of your sources.
But this dynamic isn't just a feature of the net, and it isn't new: It was always thus! Conventional wisdom thought the earth to be flat; and some crackpot or genius declared it to be spherical. Conventional wisdom thought that the speed of light was constant and that time flowed uniformly everywhere; then some crackpot or genius decided that time flowed at different rates, depending upon how fast you're moving. Conventional wisdom thought the ratio of the circumference of a circle to its diameter was an irrational number whose first ten digits are 3.14159265. In 1897, some crackpot or genius caused the Indiana State Legislature to pass House Bill No. 246, a law declaring pi to be exactly the rational fraction 22/7. (To pay homage to Timo, the difference between 3.14159265 and 22/7 "may seem to be insignificant or even non-existent; it is imperceptible to many.")
In each case, how do you, at the time, without the benefit of hindsight, determine whether the individual is a crackpot or a genius?
Consider linear and nonlinear coding; consider the situation involving Timo and me. How do you decide who's the crackpot and who's credible? In my opinion, Timo is a crackpot. I suppose that in Timo's opinion, I'm a crackpot!
Timo is - as far as I can tell - unknown in the imaging industry. But that is not necessarily any indication that he's a crackpot! Einstein was considered by many of his own era to be a kook! If I were in Timo's shoes, I would try to find a small number of independent, credible technical sources to review my assertions, and make public statements. So far, Timo has not delivered any independent, credible third party who supports his view. He was, when I last talked with him, taking his case to a pair of Finnish professors, to invite them to take a look. I hope to hear from him soon concerning their evaluations. Meanwhile, from my end, the story is suspended. (When I say "talked to him," I mean that literally - I phoned him in Finland. E-mail, web pages, and worst of all Usenet news postings, run the risk of depersonalization. We are liable to forget that we are talking to, or about, real people - humans that have not only brains, but also hearts and souls.)
Timo is apparently unable to come up with anyone in the imaging industry - or perhaps anyone in any industry - who'll step up with him and defend his view. I'm almost ready to suggest that you read about Timo's links page, and contact the author of each document listed there, particularly the rather definitive pages: CGSD, Colin Ware, and Robert Berger. I suspect that each of these people would readily support the Gamma FAQ, and repudiate Timo's view. But upon contemplating this, I really don't want to waste the time of these people. (And, I really don't want to waste any more of my time.)
I've presented the background of the story, and my philosophy concerning the credibility of scientific information (and of Usenet postings and web pages). Concerning the technical details of image coding, I recommend that you read my GammaFAQ and ColorFAQ.. Harvest Timo's "links" page, by all means. For the moment, though, I recommend that you treat with caution any of Timo's "calibration" information. Some of the pages are quite interesting, some are even quite good! But if you are not intimate with the theory and practice of image coding theory, you won't know which parts to trust and which parts to repudiate.
You may now wish to delve into the specific technical arguments on this issue. You can do this in one, two, or three ways.
Ironically, as the initial link on his links page (at 1997-03-01), Timo cites the Bourgin FAQ that was discredited three years ago! This link is annotated by Timo as "A very good FAQ about color spaces and gamma." I guess Timo's intention here was to throw the gauntlet down in front of me. (You can access a page summarizing the errors in Bourgin's FAQ.)
One final humourous tidbit: I regularly find, on the net, archaic copies of Bourgin's FAQ. My practice is to inform the hosting site of the obsolete nature of that document. Invariably, the hosting site has been happy to replace the obsolete reference by a link to the Gamma FAQ and Color FAQ. But Timo links to an archaic copy of Bourgin's FAQ, on a server located in Brazil - a server with no e-mail contact information, and no home page! Timo has managed to link to a web site that is incommunicado!
See also: Timo's technical argument - Timo and linear coding - Linear and nonlinear coding - Gamma FAQ
Copyright © 1998-07-27b