This same discussion is going on at meta, so I'll probably have to repeat everything I say here over there. The page is:
http://meta.wikipedia.org/wiki/WikiImages.org
Jimmy Wales wrote:
Daniel Mayer wrote:
Images should be a resource that are easily usable by all Wikimedia projects and language versions. All one should have to do is upload an image once and then link to it from any Wikimedia project or language version. That is why images.wikimedia.org makes sense.
Yes, perhaps. I agree with the spirit of what you're saying here, but I think that there's a bit of complexity, too.
Let's say we have a picture on wikipedia, let's say it's a photo of an elephant at the San Diego zoo. It's used on wikipedia as an illustration for the San Diego zoo article.
I am writing my 3rd grade science textbook, and I link to that same picture. In the text, I write about that picture. "As you can see from the photo, elephants love to throw dirt on their backs with their trunks."
But later someone comes along and changes the photo in the San Diego zoo article. It's still an elephant at the San Diego zoo, but a different one. And no dirt is being thrown.
So, I think that when different articles or different books use the same images, they *each* need to make a local copy.
I disagree. I think the elephant throwing dirt on its back should be called [[Image:Elephant throwing dirt on its back.jpg]], and that the elephant doing nothing should be called [[Image:Elephant doing nothing.jpg]]. Of course that means it takes longer to type the names, but I think there's other advantages to this proposal that outweigh this disadvantage.
If the image namespaces are merged (the terminology I favoured on meta since it ignores storage details), then the following policy should be put in place. In improving an image, only scaling, compression, contrast adjustment, etc. should be allowed. One name equals one click of the shutter button. If a new, better photo comes along, it should be given a different name, and the articles which refer to it should be updated appropriately. Diagrams are a bit more murky, but the policy should be that diagrams should not be changed significantly.
So, for images, I think it makes a lot of sense to have images.wikipedia.org images.wikibooks.org
and so on.
The same problem exists for various languages, of course.
I'm just saying that "available for use by any Wikimedia project" doesn't necessarily imply that the images need to be all put onto one big server.
Actually, a very sensible thing to do, I should think, would be to eventually have images.wikipedia.org, for live images used on the encyclopedia, images.wikibooks.org, etc. but ALSO to have images.wikimedia.org.
And images.wikimedia.org would be a project unto itself, so to speak, where images are collected (perhaps automatically) and copied from the other projects, and given neutral captions, and entered into a big image library. The rules for inclusion in *this* would be very strict, not relying on fair use. All images would have to be public domain or GNU FDL or something compatible.
And then if people needed an image for something, they could just copy it from there.
The main problem is the difficulty (with the current software) of handling large numbers of images, multiplied by large numbers of languages. If a better UI can be designed, allowing images to be copied in bulk to a large number of languages, then perhaps this would be ideal.
I have also suggested that perhaps images shouldn't be copied from language to language, they should be "symlinked", to use the unix term. But that scheme isn't ideal either, for the reasons listed on meta. So I don't know what the best way is.
-- Tim Starling.