Slashdot Mirror
Destroying The Myth Of The Web-Safe Palette
curmudgeon42 writes: "The folks at Webmonkey have developed a new test of the Web-safe color pallette. The results of their experiment suggest that there are only 22 colors that work across all browsers, platforms, and color depths. The article also includes a good explanation of how the different color depths operate, and some interesting strategies for dealing with the greatly diminished amount of Web-safe colors." The authors are both senior designers at Razorfish. You might not guess it from visiting some of the worst sites on the Web, but some designers are both interested in making their pages look good to all (read "most") users, and in avoiding the problems of relying on proprietary plug-ins. If your words, artwork or photographs end up on the Web, you should read it.
This gets really frustrating. I used the GIMP to design an old home page of mine using "Web-safe" GIFs. Then I took a look at it on a really nice monitor, and the colors were just horrible.
Web Monkey has been, and continues to be, an excellent resource not only on web development at the code/backend level, but also with aspects to good design practices. Unfortunately just because you can write some wicked perl scripts for a web site, does not mean you can design a pretty interface to them. It's much easier for a graphic designer to learn how to create a web page than it is for a coder to learn good design principles. Web Monkey has a nice blend to help both groups out.
Plus, they also have a great name. =)
{justin.filip | jfilip AT gmail DOT com} {http://jfilip.ca/}
Prices...
;)
Die sizes on your CPUs...
Prices...
latency...
Prices...
but the colors you can use?!??!?!
c'mon!
and just when you thought it was safe to view the web in all 216-color glory, too
Stop over-analyzing your analizations
The real problem here is the customers who insist upon complete control. They think "publishing," and somehow think that they have ultimate control over all things, and thus ultimate control over how it is presented to the viewer.
The web is a different medium. You don't take radio rules and apply them to TV, and vice versa. What works well for glossy color magazines won't work well for an indie newspaper.
I'm still fighting battles with folks. The latest here is the use of the corporate logo. The brand-identity weenies complain that there has to be one inch of whitespace around the logo, and the logo can not appear any smaller than certain dimensions, and it has to appear in the correct colors.
One inch of white space? Sure, on what size monitor?
Surely most everyone using the internet with graphics enabled has realised that web sites WILL LOOK TERRIBLE if you use less than quite a lot of colors. If you wanna browse in four colors you can, but why should my experience be spoiled by whiny pathetic ten year old technology huggers? I like gradients and drop shadows and textures. A site doesn't need to be slow to be pretty, so why should it be ugly to everyone, rather than just people who are crippled by their browser/Accelerator/whatever. Inequality in Browsers doesn't mean i should have to suffer just 'cos you do.
*serving suggestion
Unfortunately, with Netscape 4.61 on an HP-UX TrueColor display (visual), 7 of the 22 really safe colors display GIF-BGCOLOR mismatches. Of course, some of the supposedly non-safe colors may work OK on my display, but if you're trying to be truly cross-platform, the number is reduced to 15 safe colors.
What really bugs me recently is not color mismatches, but sites which have some sort of horizontal bar with many repeating vertical color streaks. It looks really ugly and I've seen it on a number of sites, so it seems to be more of a browser problem. Perhaps it's CSS that Netscape 4.x doesn't understand correctly?
Your right to not believe: Americans United for Separation of Church and
OK, this is somewhat off-topic, but it's a good story and it's sort of pertinent.
I was on a Boston to New York shuttle flight that gets stuck on the runway for 3 hours with no explanation. Worse, I'm sitting in front of three idiot consultants from Razorfish who spend the whole time talking loudly and incessantly. Remarkably, not one word of it resembled any productive activity in the slightest. "So, I conducted a series of group discussion sessions to quantify how they establish their procedures." "But, Bianca, how did you formulate the framework for evaluating their paradigms?" I was thinking back to the Slashdot article where a client sued Razorfish for delivering a shoddy site and wondered whether these clowns had worked that project.
My favorite line - Bianca is irate because a client asked her for some concrete bit of information: "Can you believe that? Hello? I'm an Information Architect, not a Knowledge Engineer!"
---------
Bollocks. What about the 6-7-6 palette that was a common "standard" for GIFs back in the old days? IIRC, the palette choice for "undithered" colours was made fairly arbitrarily by Netscape, and was not based on what the application could allocate at run-time.
More pish. They already said that the palette for 256-colour displays was drawn from a pool of 16,776,216 colour.
Well, I used to run "netscape -install" on my 8bit X server and that's exactly what it did.
If the authors felt the need to dumb-down the technical side, they could at least get it right. The article is otherwise very intersting and informative.
Be careful. People in masks cannot be trusted.
0 - black
1 - white
Reminds me of an old TV technician joke: What does NTSC stand for?
Never Twice the Same Color (prob. referring to the inevitable drift of a analog tint control)
---
They already said that the palette for 256-colour displays was drawn from a pool of 16,776,216 colour.
Actually the 'classic' VGA palette is 256 out of 262144 (6 bits per color).
When True Color cards came out, ATI came up with CoDe (color depth Extension) which was truly 256/2^24. Others soon followed.
So it's even worse: you got the 256 color drivers that support 8bpp palettes and the older ones that only go to 6bpp.
---
Besides the web-safety limitations of color selections, web designers should also be cognizant of what visitors who are color-blind will see (a subject near and dear to my heart). Webtechniques has a great article on this subject. Particularly interesting is their description of how to simulate color-blindness in order to view your own design efforts.
"If I have seen further than other men, it is by stepping on their glasses." - Michael Swaine
The authors of this article don't seem to realize that 16-bit color is 15-bit color. As a brief primer:
Similarly, 15/16-bit color is three 5-bit channels and a 1-bit padding/alpha channel, yielding 32768 colors, not 65536.
This (and some other inaccuracies in the article) cast some doubt as to how much the authors really understood what they were saying. For example, the web-safe palette still does protect you from dithering, and that's important.
Between the authors and Webmonkey, you'd think someone involved would know how to format a link properly!
---------
I hate whoever invented this stupid thing. Yes, it allows you to make sure some colors match up on low-end displays. It made sense when most of the world was browsing at 256 colors. But now, the COMMON CASE is a high-color display without color dithering! If you think minor color mis-matches look bad, take any photograph and dither it to the "web-safe" palette in photoshop or the gimp. Horror! I've seen far too many web sites which dither everything to the web-safe palette, the designer thinking that it will make everything look "right" on all displays. This article shows that it (except a very small subset) doesn't even make everything look the SAME on all displays.
Designers have a hard time learning new tools and techniques (ever seen a web site designed by a designer who does classic media?)... this habit is one of the worst.
The part that bugs me the most is the HUGE amount of time wasted on these battles. Just when I finally win an argument (and get to ignore the Brand Identity guidelines), I get socked with another weenie insisting that all text absolutely must be half-inch Garamond.
It's not that I hate fighting these battles per se, it's that I keep having to fight them, over and over and over again, regardless of how many times I try to put my foot down.
I have no
I tried for quite some time to have my little homepage to look nice. It was a pain - decent graphics took too long, then you had to mess around with different colors on the background. Then you had to go to a different machine to look at it, to make sure that it look the same on a Win9x machine w/ IE as on my Linux machine running Netscape 4.x. And, of course, it never did.
So I stopped - I think I now have three graphics on my whole "home site". I made all the colors the default (usually black on white, but on my Linux machine it's black on light gray, thanks to Netscape). So, since I made that decision, I decided to do something else on my site, to make sure that people would come back. Something rarely seen on the Web today (though that hasn't always been the case). I put in actual content. Naturally, it's my website, so it's all about me and my not-so-exciting life, but still - there is actual content on my web site.
- mikeh
Is there something wrong with my Web Safety? Should I upgrade my version of Lynx to get all 22 colors?
Carefree highway, let me slip away on you.
If you want to be truly safe, you need enough contrast between your background and your foreground so that it works on machines with 1-bit colour.
But why bother? Nobody would seriously use such a machine to browse the web, as useful as it could be for other things. Similarly, do you really need to appeal to users with only 8-bit colour? I mean, so long as they can see something, is it really worth making it look perfect?
Oceania has always been at war with Eastasia.
Gamma kinda goes like this, perception of differences isn't based on an absolute difference, but a percentage different. I can feel the difference between one ounce and two ounces much better than I can feel it between 30 punds and 30 pounds 1 ounce. The absolute difference is still 1 ounce, but the percentage is radically different. The perception curve is based on an exponential, and that exponent is named gamma.
The percieved color difference between 0x00 and 0x33 is radically different between 0xCC and 0xFF. You actually want a perceptually equidistant color space, not mathematically. Ever wonder why dark gifs look so bad? because there is too much spacing (perceptually) between colors at the bottom end.
BS, programmers who don't understand color theory or are too lazy to program it right liked the mathematical simplicity.
I've written low-level drawing code that had to work on 3 dozen video cards, so I know entirely too much about this. There are actually 3 flavors of "high color":
* 15 bit, no alpha channel, aka "555" color with 5 bits each of R, G, and B.
* 15 bit with alpha, aka "1555" with 1 alpha bit and 5 bits each of R, G, and B.
* 16 bit which (on PC hardware at least) is always "565" color, with 5 bits each for red and blue and 6 for green (because the human eye is more sensitive to variations in green).
You can of course treat 555 and 1555 the same in most cases. Older boards tended to be 555 format, while most newer designs are 565.
I recently found time to begin learning photoshop for the first time at work. I thought I'd redesign my web page with my newfound skills.
What I found quickly was that even using a web safe palette in photoshop, and then viewing my page in netscape on the SAME MACHINE, the color was skewed. (Curiously enough, the color in question was actually closer to what I'd intended in netscape on a different (Mac) OS).
It didn't take me long to decide that there was pretty much no hope of complete color concordance, and I gave up, and just tried to make something that would look decent even if the colors drifted. That's pretty much what you have to do. Sort of like computer security?....
Tweet, tweet.
In the article, the authors claim that when your monitor is in 15-bit mode, there are 32,768 colors it can display, and that these are chosen uniformly in a 32x32x32 quantization of RGB-space. This is correct. However, they also claim that none of these colors exist in true-color mode (except black and white and other pure colors) because, for example:
In 15-bit mode, the color (1,1,1) where each number is in the range 0-31 gives us about 3.23% gray.
In 24-bit mode, the closest colors you seem to have are (8,8,8) and (9,9,9) in the range 0-255 which correspond to 3.13% gray and 3.53% gray.
However, this assumes that at a hardware level, there is a difference in the signal being sent to the monitor between 3.23% and 3.13% for each color channel. Is that really the case? My guess would be that when you're in high color (15-bit) mode, each pixel gets translated to its nearest 24-bit equivalent inside of the video card before the signal sent to the monitor. This is almost certainly the case when the connection to the monitor is digital, like in some new flat-panel displays. Anybody know about this for sure?
I'm glad to see that they addressed the concept that different rules apply to different media.
My favorite is:
STRATEGY 9 -- Go back to print design.
Anyway, I ran into this trouble a couple weeks ago. What I _REALLY_ hate is if you have a GIF that's a solid color (say #006000) and you set the BGCOLOR of a table or body to the EXACT SAME COLOR, the browser displays the GIF and the BG differently. Drives me nuts.
Some day there will be a browser that actually works. Some day a long time after that, most people will be using a browser that actually works.
Until then, I'm forced to send design back to the design team over and over until they get a clue.
Someone else hire me.
Yet another inaccuracy in the article:
:-P
These "senior" whatever-they-are's didn't bother to look closely at their tests. They say that the color mismatches occur in high-color modes because the browser has to pick between (as an example) 1.9 and 2.1, from an original 2.0, and it doesn't always pick the same thing...
If they would look at their example images (ghost.gif and obvious.gif, linked from within the article), they would see that the BGCOLOR for the table cell was solid, but the GIF was *dithered*. They claim that it is a bug in the browser. You could consider it a bug, I suppose, but it's really just the fact that the browser assumes that BGCOLORS should be solid, and thus picks the nearest color, whereas images are quite often *not* solid shades of color, and they usually benefit from dithering.
Another nitpick: The fact that they claim these colors that don't pass their test in high color aren't "web safe" is inane, at best. They consider these colors "unsafe" because they are shifted slightly in high-color? Um, how often can you get colors to display consistently across all sets of hardware/software? Never. A little color shifting is irrelevant. *Maybe* you can have an issue with high-color dithering (which they didn't seem to notice), but that's pushing it...
Duhhh, I feel special because I picked apart an article written by guys making lots of money...
The extra bit in 5-5-5 could theoretically be used for an alpha channel, however in reality it is almost always ignored.
The advantages of the 5-5-5 format is that the color components are more equal, so you get a full 32 shades of true grey rather than the pseudo almost-greys of the 5-6-5 format. Of course, for most photographic data you want as much color depth as possible, so the 5-6-5 format is preferred.
Download a fast DirectX Tetris Clone [276 k]
The concept of a web-space palette has nothing to do with consistency between hardware setups.
The concept of a web-space palette has everything to do with using one general but limited set of colors for all images, when the hardware has only 8bpp in which to render.
If you're using a video card in a 8bpp hardware mode (common in 1994), the hardware must use a palette: a lookup from 8bpp indices to the analog RGB triple. There is only ONE hardware lookup table for all applications that have access to the screen.
In such a scenario, the graphics manager (Windows GDI, the MacOS equivalent, or whatever other system) generally gave the foreground app priority. It could load as many colors into the hardware lookup as it wanted (up to the obvious 256). All other apps had to use a "logical palette", which was a wishlist. The manager would map any spares to the best colors from all background apps' wishlists, and map any other graphics to the nearest equivalent in the definitive lookup.
Changing the hardware lookup meant "palette flashes", most commonly seen when switching between two graphics applications, each of which trying to optimally select colors for its graphics documents. The flash was because the hardware lookup changes took place instantaneously, while the software had to refresh their images at whatever speed the CPU and video memory accesses allowed for.
A web browser, unlike a graphics app, is used to render many images at once. Thus, it must in turn emulate the SAME sort of wishlist strategy for each image, not just each application.
The 216 color "safe" palette was proposed, because this was the most theoretically evenly spaced set of RGB values that filled the whole RGB gamut (6 levels R, 6 levels G, 6 levels B, 6*6*6=216, 216<256).
If the web browser app registered the 216 colors (plus any common shades of gray used by the default OS GUI), then it always had *some* chance of making a reasonable nearest-color compromise in every image it was asked to draw.
There is NO way to make 100% match between different mediums. Two different digital-to-analog chips (DACs) made from the same wafer of silicon will still have different thermal and amp response curves. The best you can do is approximate.
Macs and SGIs have inherently focused more on color reproduction quality, but they still vary a lot. PC cards are more interested in pixel pushing than in DAC quality, but they're better than ever at making good color. The reputations stick: Mac goes for high-saturation color, while PC goes for a flatter gamma.
That's not even getting into RGB vs CMYK color spaces... the monitor, even in "true color" 24bpp modes, can only approximate about half of the actual color space available to the human eye. Print media can also get only about two thirds.
Good logo designers have to consider embroidery, silkscreening on fabric, silkscreening on plastics, diecut metals, print and onscreen uses for that logo. You think the web was a limited environment for color choices!
In short: "Safe palettes" are good for reducing compromised color selections, and a common palette from app to app helps in reducing that hardware palette lookup flashing. If you're in a higher color mode, you don't get any of the latter case, and the 216 colors are irrelevant to the compromises made in the former case.
But you STILL won't look like your neighbor.
[
OK, so when doing web graphics, I've got two choices...
1) Use a low-color palette and have graphics and pages that look generally crappy to everyone, or
2) Stick with the "don't worry about it" method and have graphics and pages that only look crappy to those with amazingly low color palettes.
-S
--- What parts of "shall make no law", "shall not be infringed", and "shall not be violated" don't you understand?
Am I the only one who thinks this is a non-issue created by people obsessed over aesthetics and graphic design? I have never once come accross a site whose color combination has limited the functionality of the site. Sure, if you have photorealistic images or something, they might look slightly different, but really, for most cases, does precise matching of color matter *that* much (not to slight the color blind). I mean, the BSD and YRO sections of Slashdot are pretty damn ugly, but somehow I still manager to cope. Are there really people out there who are going "Hey, this site looks subtly different on my Mac! Damn you! Burn in hell! I will never buy your products!".
Who really cares if the colors are a bit off? (and I understand if it's a matter of principle - you don't design poorly if you know there is a right way to do something, but still...this seems like splitting hairs)
It's 10 PM. Do you know if you're un-American?
As a scientist, I give presentations occaisionally, and a (male) audience member later told me he was red-green color-blind, as was ~10% of the male population.
Since then, I've tried to never use both red & green as the only distinguising characteristic on a chart, etc.
While I'm at it - blue & red should not be placed next to each other, generally. Since they fall roughly at opposite ends of the visible spectrum, the eye's focal power differs the most between those colors. As your eye/brain tries to focus properly on two colors that require slightly different adaptations, you can perceive a "vibration" -- the boundary between the red & blue will have a high-frequency shimmering or vibrating appearance.
This is not universal, and is most noticeable between bright solids with adjacent, straight edges.
Further OT - it can also be used to interesting effect. There's a laser-tag place nearby, and the carpet has a blue-grid pattern offset on a red-grid pattern, illuminated partly by blacklights. From the observation gallery it has a 3D effect, with the carpet looking as if it has 12" deep holes in it. I believe it's due to the red-blue focusing issue.
-----
D. Fischer
ShoutingMan.com
There's evidence to back up your statement: Statmarket.com, a subscription service that provides browser stats based on samples culled from sites that user their server stats technology, used to be free about a year ago. Even back then, their stats seemed to indicate that the prevalent setup for machines was 800*600 and 16-bit colour.
They say, "Unless you work entirely in black and white, approximately half your general audience won't properly see the colors you select for your site."
Ahem. Let's rephrase that: "No one but you will properly see the colors you select for your site." If they think that slight color shifts are a problem, they need to standardize everyone else's: browsers, OSs, graphics cards, monitors, viewing conditions, and eyes. Enjoy.
Forget True vs High color. There are more basic issues at work:
1) Non-color devices: Palms, cell-phones, terminals, lynx, etc.
2) The reason people care about "websafe" colors is that they want the client to see what the designer designed. But if I adjust the settings on my end, I don't see it anyway. "The settings on my end" include everything from constrast/brightness/etc on my monitor to the individual color tweaks available on some TVs (as in "WebTV").
It is literally impossible (not just difficult) to make this work, so why not design around it? Stop making pretty colors cover the fact that you have no content and actually give me some meaningful information.
--
Linux MAPI Server!
http://www.openone.com/software/MailOne/
Linux MAPI Server!
http://www.openone.com/software/MailOne/
(Exchange Migration HOWTO coming soon)
Old-timey graphic designer motto (which isn't taught in schools anymore, to judge by Wired and it's ilk):
I'm kinda old-fogey about this. If it's black, you read it. If it's blue, you click on it. If it's grey, it's the background.
Jeez... If you're site is all about matching colors and transparent GIFs, you've got a brochureware site. Don't sweat it -- people will look at it once and never come back.
(Browsing Slashdot in "simple HTML mode"...)
Potato chips are a by-yourself food.
One of the most interesting things I learned in this article is how inconsistent web browsers are when attempting to render colors in 15 or 16-bit modes. I can imagine this happens because sloppy programmers might convert from an 8-bit number to a 5-bit number by doing a bit-shift, incorrectly ignoring the less significant bits.
This is a big deal, for example, if you need solid colors (like table BGCOLORS) seamlessly blending with GIF images. I can imagine this coming up sometimes, but not THAT often. Luckily they offered some suggestions to remedy this problem (like using a transparent color in your GIF where it blends with the background).
The authors of the article, however, seem to imply that one concern is that the colors people see are not the colors you intended for them to see. This is a different issue entirely! Just the fact that most monitors have brightness/contrast controls, plus the differences in gamma used by Macs and PCs, and other factors like this virtually guarantee that most users will not see exactly the color you intended.
You would NEVER, for example, expect to do business with someone whom you pick up in a white limo in China - White is a color reserved for funerals, and typically associated with death.
Also, keep in mind the industry you are dealing with. For example, in Nuclear circles, where many people started out in the Navy, red means ON(hot) and green means OFF(cold)... So a flashing red marker might get misinterpreted.
But then again, developing a web site with your audience in mind is common sense, isn't it? Well, isn't it?
The REAL jabber has the /. user id: 13196
The REAL jabber has the user id: 13196
What you do today will cost you a day of your life
A secretary in my office has a windows95 machine that the sysadmin never installed the correct video driver on. Her machine is set on "default display" which only displays 256 colors.
It never bothered her that everything looked really weird.
Consider your audience.
Design accordingly.
Roughly 99.9% of the general population are not anal retentive web developers.
"tension is the great integrity" -R Buckminster Fuller
I know your exact problem. Photoshop 5.5, right? Go to the RGB settings under Color Settings under the File menu. Change it to Monitor RGB or Adobe RGB 1998. Also turn off the embedding of a profile for RGB images.
Photoshop 5.5 has a lot of nice stuff, but fungling with my colors is a big bugaboo.
Potato chips are a by-yourself food.
We had a client who was checking out their website on two different (same brand, model etc) laptops, same OS, same browser. Same color depth.
They couldn't figure out why the colors weren't exactly the same, and over the phone, we didn't catch the problem either. Naturally, they thought it was a problem with our design or programming.
So, when they were here for a meeting, they brought the laptops along, connected to the site, and said "There, see? The colors aren't the same."
They were right. The colors weren't the same despite exact same configurations on the two machines.
I _REALLY_ impressed them when I reached up and adjusted the brightness and contrast of the darker laptop.
[eyeroll]
I read just enough color theory to survive my Ph.D. prelim several years ago, and promptly purged it from memory, so take this for what it's worth...
:)
:)
You're right, human perception is not a linear function. We perceive green most strongly, over red and blue (in terms of a fundamental RGB colorspace). Also, our perception efficiency is roughly a bell-curve for each spectrum (R, G, B) with the max at some particular hue of each one.
Our perception is also significantly affected by external conditions: adjacent colors, lighting.
Two adjacent (non-white) colors will be perceived differently than if isolated on a white background.
All lights have a spectrum; put another way, few (none?) consumer & commerical lights produce "white" light. Mercury-vapor lamps (i.e. street lights) are distinctly yellow. Neon lights, I think, have more blue than sunlight. So on. This leads to the common experience of buying a purple shirt in a neon-lit store and finding out it's blue in the sunlight (or whatever; I don't recall the particular color shifts). Likewise, a certain blue, say, may be very appealing, but when surrounded by a certain green, say, it no longer appears as expected. Combine that with the lighting issues, and you've got a mess.
Related to lighting is the eye's light-adaptation level. When dark-adapted, nearly all color vision is lost, with the eye being most sensitive to red. So, a someone reading a magazine (or computer monitor) at night with dim lighting will see the colors differently than outside on a sunny day. (aside: I usually sleep on my side, so one eye is covered by the pillow and the other isn't. On a sunny morning, the non-covered eye is somewhat light-adapted whereas the other is more dark-adapted when I wake up. My green-lettered alarm clock then appears dual-colored. One eye sees the green but the other sees more yellow. Try it sometime: keep one eye covered for about 10 minutes and then look at strongly colored objects
Back to the original poster...uniformly spaced color doesn't seem like the best choice for human interface issues. But since when have computer programmers concerned themselves with human interface issues?
-----
D. Fischer
ShoutingMan.com
Good graphic design is better than no graphic design.
Useful != plain.
Useful != ugly.
Well designed sites include, but are not limited to, brochure sites.
Default background with text at the full width IS a design choice, and usually a bad one.
You can have a useful, well-organized site that people return to, AND doesn't look like it was designed by programmers who spend 14 hour days in their caves reading nothing but man pages.
* And remember, it's spelled N-e-t-s-c-a-p-e, but it's pronounced "Mozilla."
take web color choice advice from the people who built this. Right...
Cypherpunks: Civil Liberty Through Complex Mathematics. Those who live by the sword die by the arrow.
Comment removed based on user account deletion
And that's why on *expensive* graphics cards, there used to be a pantone-colour-matching hardware/software combination.
IIRC, on the high-end #9GXE cards (around 1993) you can get this thing that will measure the colour saturation and whatnot, and will send adjustment values, until it gets it right. So, Green xxxx.x will be Green xxxx.x on screen too.
--
The 22 colors they've found are ones that not only can be displayed without dithering, but look exactly the same whether in an image or as an HTML background. The times you'll run into problems using colors that aren't in these 22 are when you're trying to put images on that background.
For one thing, you can still use transparency safely on top of any of the 216 colors. Otherwise, you can still use all 216 colors in the foreground of the image - only the background has to be one of the 22 colors if you want it to look right.
These 22 colors include, of course, black and white, which I'm guessing account for at least 90% of the page backgrounds on the Web.
--
No more e-mail address game - see my user info. Time for revenge.
Win dain a lotica, en vai tu ri silota
The latest generation of Softimage renderers use a format which has a 32-bit floating point number for each color value of each pixel. Why? So realistic lighting models in radiosity renderers will work. The real world has many orders of magnitude in intensity variation. High-end animation sometimes needs to work in a true intensity space, so that sunlight and shadows are properly represented. This is necessary to make compositing work right. At the very end, there's flattening as the images are rendered for display, but in the middle of the process, you need that kind of dynamic range.
If you are doing design in-house, that may be alright. But if you have been hired to design for someone else, your real audience is that person that hired you. If she thinks it looks like crap, then it does, regardless of what the world will think.
If you make your living off contract design, your audience is completely different.
wishus
Vote for freedom!
---
And I wonder why I haven't got any hits in the past six months... (Yes, I use frames, but it is lynx friendly.)
I'm trying to teach myself to set people on fire with my mind... Is it hot in here?
Actually - there are gamma conversions somewhere between the equally spaced hex numbers (which have to be equally spaced, obviously, they're integer values) and the intensity displayed on your screen.
Unfortunately, this is not the same Gamma between platforms...
I believe there's actually a gamma conversion in your video card and then another gamma conversion in your monitor, and then on some monitors the 'contrast' knob actually adjusts the gamma, but on others it's really contrast... I'm kinda surprised, actually, that they managed to find -any- websafe colors -at all- given all the variables.
Anyway, in theory, PNG is supposed to solve all these problems (at least for images), by allowing a gamma value to be attached to the image so the appropriate adjustment can be made, but no web browsers are supporting the gamma-functions of PNG, or they weren't when O'Reilly's PNG book was written.
--Parity
--Parity
'Card carrying' member of the EFF.
Something they didn't really touch on is the fact that different OS, Browsers, Video Card Drivers and hardware will give different results. Case in point, a buddy of mine made a nice grey gradient for the background on his webpage. Easy on the eyes at 24bit. Bad banding on my old Matrox card at 16bit. We checked it on an old Diamond Stealth card at 16bit and it looked horrible with uneven bands with shades of green and purple in them.
Of course, it all comes down to what half of the replies to this article have said, the web is not print publishing, you cannot have pixel perfect designs for 100% of the audience. Websites that are built for standards compliance and accessibility first and eye candy second will always be more robust in face of alternative browsers and browsing platforms.
Bleh!
Never Twice the Same Color (prob. referring to the inevitable drift of a analog tint control)
Although drifing off-topic, but still very geek, I've always heard it as "Never The Same Color", which rolls off the tongue better.
DeepGeekStory: Most people have heard of the Rocky Horror Picture Show, where people get dressed up, throw rice and toast, and yell lines at the screen. Fewer have heard of the little known sequal (sorta) titled "Shock Treatment". People also get dressed up as characters in Shock Treatment, act out the parts, and yell lines.
One song (Looking for Trade) is filmed under red light with red walls, with white spots of color. A friend yells the line just before the song starts: "Hey, Show us your complete disregard for the NTSC format!!". Of course, on televisions (including projectors, which is how we were watching the movie), the red smears and blurs across the screen because of NTSC's lousy color signal handling.
ObObservation: Could you ever have a NTSC colour signal? Or would be oxymoronic, like a PAL color signal?
--
Evan
"$30 for the One True Ring. $10 each additional ring!" -- JRR "Bob" Tolkien
First of all, you're completely missing the methodology of the testing procedure (as are many other Slashdot readers). You're also assuming that they displayed the color on two separate machines simultaneously with different versions of a browser and just eyeballed 'em to see if they looked the same. This could not be further from the truth. From the article:You can look at the te st page to see what they're referring to (I'd suggest dropping down to 8-bit/256-color to get the point).
It has nothing to do with colors not looking quite right from one videocard/monitor/OS to another, it's all about the way the browser/OS match colors. That is, if you tell the browser to display #CCFF66 in 8-bit mode, it has to find the closest match that's actually in its available palette. Here is the problem: some browsers might determine that the closest match is #CCFFFF and some might determine that the closest match is #CCFF00.. and then when you throw in a GIF that's supposed to be the same color, your computers tends to use a completely different "algorithm" for determining how to match the GIF's color. So, the actual internal HEX value associated with a single-color GIF that's #CCFF66 and a section of a page with BGCOLOR set to #CCFF66 might not be the same on two different browsers. It is a tangible and measurable difference in values.
This isn't some sort of video-dependent thing or weird, quirky phenomenon that just makes colors "kinda look different." There is a qualitative variation in the way that colors are selected, regardless of how similar (or different) they might appear, depending on your monitor.
Although you could test more browsers and operating systems (this test was only concerned with Windows and Mac), there is nothing arbitrary about these 22 colors.
In fact, JPEG relies on perceptual coding, so that the perceived effect of the lossy algorithm is minimized.
The exceptions to the rule are SGI machines (which were designed with ray-tracing in mind, so they assume physical coding in the frame buffer and use gamma correction to convert to perceptual coding) and Macs (which have a gamma correction intended to mimic the dot gain of the original Laser Writer.)
For more than you ever wanted to know about gamma, see Charles Poynton's Gamma FAQ and the sRGB proposal.
Wish I could claim some kind of freak genetic abnormality that allows me to remember obscure stuff, but in reality, I fought with it for over a week trying to get Photoshop 5.5 to work as well as 5.0 did for me. It's hard to forget such pain.
Potato chips are a by-yourself food.
That's exactly what the friendly folk at Microsoft are trying to do.
Whoa, put away the flamethrower! I did indeed read the article. If I made a couple of mistakes, it doesn't mean I'm "off the charts." Because I wasn't entirely in error. Let's assume, for now, that it does indeed have nothing to do with hardware (caveat: keep reading). Even so, the 22 colors under discussion might not be right in your environment, because how many of us are using one of those browsers (NN, IE) on one of those platforms (Win, Mac)? Lots, but probably not all. So the color matching on your chosen platform (Linux, and HP like the parent post said) might be different yet again. Even without taking hardware into account, there are tons of variations that the article didn't account for. So, there's essentially no web-safe palette, at all. You could probably scope out your demographics, discover that ~50% of people are using some IE variant on win9x, and create a "web-safe-enough" palette.
But, if hardware makes no difference, then what about this (from page 5 of the article):
Admittedly, we got different results in what would seem to be the same conditions: one Win9x machine would fail for a color, and another such machine wouldn't. We decided to play it safe. Really safe. We found that only 22 of the 216 colors we began with did not end up being shifted incorrectly in at least one viewing environment...
And then they linked to the "reallysafe palette" from there. Why would windows machines with (presumably) similar software setups behave differently? Isn't it possible that it's somewhere in the video driver, and therefore at least partly hardware dependent?
Apparently you didn't read the article. The biggest incompatibility is trying to present colors the same to both High Color(15 and 16 bit) and True Color(24 bit). Even if you leave the old 8-bit browsers out of the picture, you have to contend with the fact that there is no compatibility between High and True Color. There is no overlap: if a color is represented in True Color, by mathematical definition it is not represented in High Color.
Seeing as how ~56 percent of people use High Color and ~38 percent use True Color, you've got a real problem there. You cannot present the same high-quality image to both audiences. Getting rid of High Color settings would largely solve the problem, but unfortunately, these people are in the majority.
No, Thursday's out. How about never - is never good for you?
Can you imagine the razzle-dazzle it did on NTSC???
As soon as I saw 'webmonkey' and 'Razorfish' I knew where this was going. I've got to join the chorus singing "design for standards, usability, and accessability" and just nod my head and smile at the pixel perfectionists. As I've said for years, the Web is NOT a book or magazine, and treating it like some jumped-up animated version of said media is a dead end.
Cases in point:
When the photographic process was invented, it was treated as just a new kind of painting or drawing. Fortunately there were experimental people who realized that it was a new medium with new rules.
When moving pictures were first invented, early films were made of stage plays. One camera, mounted in front of the center of the stage, never moving. Thankfully directors and cinematographers figured out rather quickly that this is horribly boring.
25 years from now we'll be saying that when the web was invented people treated it like a book or magazine, or maybe a little like TV. I predict that 25 years from now we'll consider these concepts quaint early forays by curmudgeonly glue-stained and Xacto-nicked designers.
Your statement about perceiving green most strongly is not quite true. We tend to perceive high inensity red strongest, followed by most of green and then blue. If you look at real landscapes, the sky blues are quite constant compared to green plants but small amoutns of red stand out. I seem to remember seeing where out blue sensitvity is something like 1/3 to 1/10 of the red/green while red and green are about the same but have different gamma curves.
The plot of brightness is actually a curve with a function "(x/255)^G+K" where G is about 2.3 for your normal crt (K is the brightness of the dark screen).
It just so happens that this curve very closely matches the human eye's color sensitivity. Thus each different numerical value results in approximately equal differences in perceived color.
LCD screens do not naturally have this response, but the manufacturers have been forced to copy it to make displays designed for CRTs look correct.
A big problem with CG is "partially clued" people that are aware of the screens non-linear response, but think that linear response will result in better pictures. They adjust the software or the D/A converters for this and actually produces worse output, where there are noticable steps in the black and many indecipherable values crammed up in the white end.
PC manufacturers, being cheap, have not done this, and actually resulted in better color. Macs and SGI workstations suffer from default settings that try to "solve" this problem.
Most video cards use an 8-bit D/A converter, since they can be switched to some 8-bit (24 bit in PC-speak) mode. (some older cards use a 5 or 6-bit converter so that 8-bit mode is actually a waste of memory!) They almost certainly pad the 5 or 6-bit data with 1 or 0 to get the 8-bit input, or they might copy the top 3 bits to the bottom.
Perhaps more important is that this difference is miniscule compared to normal differences between CRTs and brightness settings and D/A errors.
-----------
"You can't shake the Devil's hand and say you're only kidding."
Yes, in Canada we spell it 'colour'.
---