Originally Posted by solipsism
I'm by no means an expert on graphics so please correct me if I'm wrong, but I am under the impression that vectored images are best when try don't contain a lot of detail and at small sizes can be significantly larger than bitmaps.
Not the size of the AI logo In the arrticle. What would be the file size for that same image, since it is simple? Is their proxessing overhead for vector over bitmap that can slow down the page render? If vector images are generally ideal why does Apple offer Icon Composer and create all it's apps as 16^2, 32^2, 128^2, 256^2, and 512^2 as bitmaps?
Originally Posted by ajmas
As you indicated, It all depends on the complexity of the image and the supported features of the vector format. Also, is the vector format binary, text, or compressed text. Any comparison would have to be done with JPG or PNG vs SVG. Then again, this presumes that the resolution of the image is suitable for all display devices and scales. Bitmaps would have to be available in multiple sizes, whereas a vector can be scaled to almost any resolution.
You see, raster vs vector was one of the big battlegrounds in web design before the whole Flash debate came to the forefront.Currently almost all sites do not incorporate vector graphics in web pages, except for Flash elements.
The only thing that could be considered "vector-based" would be background colors, and "div tag and table layouts". Yes, there is SVG, and CSS provides some vector tools. But by convention, for the most part, SVG and CSS shapes are not used (although CSS may be gradually used for eg. drop shadows and rounded corners). If you think Flash is a big deal, the SVG/CSS vector issue has been smoldering for much longer. SVG was supposed to be a saviour in all kinds of ways but it just didn't pan out. In fact PNG support became more important because it offered non-lossy compression as well as alpha channel (transparency).
In an ideal situation everything would be vector based so that it can be rendered at an extremely wide range of sizes and still look fine. Two challenges exist, mainly because the industry did not go this way.
One is the file size, as pointed out in previous posts. Bitmaps provide mostly predictable file sizes and thus can be planned for accordingly. Could intelligent compression be applied to complex vector files in the future? No doubt, and even now sometimes really complex vector files can be surprisingly small.
Two, and more significant, I think, is the rendering of vector files. Let's say for the icons on a Mac desktop. Let's say you open your Applications folder, and have 20 to 40 icons displayed, and you scaled them up and down. At this stage, on Mac or PC, rendering using bitmaps is a heck of a lot faster. Again, I don't think this is a hardware limitation. In the bitmap-based solution, the icons are usually encoded as PNG so they have to be decoded anyway, and the Finder has to resample icons on the fly and display them antialiased with transparency. So the Finder is doing some non-trivial work there already. Another example is we have fast decoding and displaying of demanding video compression eg. H.264, even before GPU acceleration came into the picture.
So if the effort was put into storing, compressing and rendering vector graphics, we could really have had everything really vector based.
I am no expert on early computing and machine/assembly code but I think all this stems from when graphics was first used in personal computers. In those situations, literally bit-mapping each pixel to one bit of data was the most efficient way of displaying complex images. Later, with 256 colours, then 64k(?) then millions of colours, it was just an extension of bit-mapping, only this time applying more bits for each pixel to store colour information. You wanted larger, clearer pictures? Just display more pixels, and give each pixel more bits of colour information. You wanted transparency for blending images over each other? Just give each pixel an additional set of bits on it's level of opacity. But it was still pixel-oriented.
So at the end of the day, only with HTML 5, newer CSS, PDF, Core Image, 3D gaming, 3D interfaces and so on, are we starting to move away from a bitmap-oriented world. But as you see with web pages, existing OS interfaces and disasters like the Wired application, we are still stuck with mapping bits.
While I'm at it, this situation is actually what led to the tremendous rise of Flash circa 2000-2008. Because you could display a lot of vectorised graphics, effects, eg. text layers, transparency, drop shadows, curves, rounded corners and it would scale well (later on bitmaps in Flash would also scale well) to a really wide range of resolutions and look good (and interaction etc would work predictably across browsers!) on almost any computer, albeit it had to have the appropriate Flash version (but each new version of Flash was suitably backwards compatible with older Flash files). Rendering of complex vector art (eg. illustrations) was pretty fast too, so that led to the rise of Flash gaming and Flash advertising... Though said rendering would be crazy sluggish on Macs, and now a big burden on mobile devices.