Correction-- per Tim Starling on Meta, the portion that we pay for (and is not donated) comes to around $300,000, making the savings much closer to $3,000-4,000 i.e. even more meaningless.

-Dan

On Mon, Jul 28, 2008 at 2:57 AM, Dan Rosenthal swatjester@gmail.com wrote:

I fail to see the benefit of this trade off when compared to the extreme negative publicity we'd get for it. The 2007 finances report says we spent approximately $389,000 on internet hosting. 1.5-2% of that is around 6,000 dollars a year.

You couldn't hope to buy off the negative publicity we'd get from this for a $6,000 a year savings.

-Dan

On Mon, Jul 28, 2008 at 2:44 AM, Joe Szilagyi szilagyi@gmail.com wrote:

...

Just a heads up that there is a discussion to see if disabling all "hotlinking" of images and media from external sites is a good idea, here:

http://meta.wikimedia.org/wiki/Meta:Babel#Proposal_to_disable_hotlinking

Tim Starling posted some statistics that indicate 1% to 2.5% of all requests are from outside sites leaching WMF bandwidth resources. In theory this could save the WMF 1% to 2.5% or more of it's bandwidth costs.

• Joe

---

foundation-l mailing list foundation-l@lists.wikimedia.org Unsubscribe: https://lists.wikimedia.org/mailman/listinfo/foundation-l

-- Dan Rosenthal

Seriously. Also, I just saw the justification for this and about dropped my jaw.

*"We are not a free image host* - *Money and server power is being wasted on hosting millions of images for websites that are completely unrelated to the Wikimedia project. Here is an example of a search for en.WP images that do not mention Wikipedia, [10]http://images.google.com/images?um=1&hl=en&safe=off&client=firefox-a&rls=org.mozilla%3Aen-US%3Aofficial&q=site%3Aupload.wikimedia.org%2Fwikipedia%2Fen+-intitle%3Awikipedia&btnG=Search+Imagesover 2 mil. We should not be using our servers for this, as I doubt anyone donated money with this in mind. Copying of free images is obviously fine, just as it is with text, but we don't (as far as I know) host the text for any other site, I don't see why we should do this for image"*

No we are not a free (as in beer) image host, but we do host free (as in beer and liberty) pictures. The vast majority of these images are not actually hotlinked anywhere, and take negligible amounts of bandwidth. On the other hand, by keeping free (as in liberty) images on our servers (preferably commons) we allow them to be used to benefit future Wikimedia projects. And the completely irrelevant ones should have some sort of deletion criteria if they are not being used. I don't see the need to block hundreds of legitimate sites from hotlinking, and swarm us with negative publicity for this point.

"*We should not host non-free and fair-use images for anyone to use without reason* - *Wikipedia hosts many non-free and fair use images for use within the project, but we can not limit usage of hosted non-free images on other sites. Ethically and legally is not a good idea to host copyrighted works for anyone to use with any purpose."*

We don't. Non-free images that are not being used, are supposed to be deleted. What people do outside of the project is their own concern if our uses are legitimate; and furthermore it's not within our scope of concern to be policing external use. Blocking hotlinks is not the solution to this problem either.

*We bog down Wikipedia with uploads of nonsense photos *Statistics show that they're less than 2% of bandwidth. $6,000 a year, tops. Ignoring the blanket statement that the photos are "nonsense" (because really, who's spent the time to check every one of these millions of photos and make a qualititative decision on them?) they don't really hurt us.

Now, lets sum up the bad reasons: 1. Miniscule benefit. Implementing technical procedures for miniscule benefit is usually a bad thing, especially if there are problems and they don't go well. See below. 2. PR - We are all about getting free content out to people. What kind of message does it send when we suddenly start restricting that content? The second this gets picked up on any major news media outlet would be the second that there is a backlash and petition against the WMF, and we lose face everytime we say "the sum of all human knowledge" or "bringing free content to the world" and get the responce "unless it's hotlinked, right?" Honestly, I wouldn't think it unforeseeable that we'd lose more money in lost small donations than we gain in saved bandwidth. 3. Hurts potential business agreements with partner organizations. With this enabled, NOBODY would be able to hotlink images from us. Not Google, One Laptop Per Child, not anyone. This potentially hurts our ability to make business deals with partner organizations. 4. Outside our scope. It's simply not our primary, or even secondary concern what other people do with our content, but rather what we do with our own content. 5. Hurts legitimate users from reusing our content to say good things about us.

Some of these concerns may be slightly overstated, but any one of them outweighs the lack of benefit from enacting this change, and in the aggregate they completely outweigh it.

-Dan

On Mon, Jul 28, 2008 at 3:01 AM, Joe Szilagyi szilagyi@gmail.com wrote:

On Sun, Jul 27, 2008 at 11:57 PM, Dan Rosenthal <swatjester@gmail.com

...

wrote:

...

I fail to see the benefit of this trade off when compared to the extreme negative publicity we'd get for it. The 2007 finances report says we

spent

...

approximately $389,000 on internet hosting. 1.5-2% of that is around 6,000 dollars a year.

You couldn't hope to buy off the negative publicity we'd get from this

for

...

a $6,000 a year savings.

-Dan

Agreed. I just saw Tim's numbers on it after mailing out the FYI here. Even if it topped at 10k, it's not worth it. Maybe in several years of scaling...

• Joe

---

foundation-l mailing list foundation-l@lists.wikimedia.org Unsubscribe: https://lists.wikimedia.org/mailman/listinfo/foundation-l

This proposal is completely wrong-headed. Wikimedia's goal is to provide free information, to everyone. If it can do this helpfully and at reasonable cost by providing free image hotlinking of images it happens to have anyway, it should do so. As for arguments about costing Wikimedia money, please see http://en.wikipedia.org/wiki/Wikipedia:Don%27t_worry_about_performance. If the sysadmins think it's a problem they'll handle it, and it's not.

If some of the images people are hot-linking are non-free, it's up to the ones hotlinking them to determine whether their particular use meets the requirements of fair use -- Wikimedia need not police third-party use of its resources unless there's a complaint. Others have a right to use copyrighted works under fair use, and Wikimedia should not stand in their way. And if people are uploading images that are useless to us so they can hotlink them (which I haven't heard is a big problem), just delete the images.

There's no need for selfishness here. Wikimedia's goal is advanced about as much if third-party sites can use its information directly as if they have to direct people to Wikimedia's sites to get it. The point is to disseminate the information, not to disseminate it with a Wikimedia-owned logo in the top left corner of the page.

Essentially the same arguments apply to watermarking and other overly protective measures. The status quo is perfectly fine with respect to third-party use of Wikimedia images. I've hotlinked Wikimedia images myself on more than one occasion.

On Mon, Jul 28, 2008 at 11:32 AM, Gregory Maxwell gmaxwell@gmail.com wrote:

Although, as it stands right now we're really encouraging unattributed use. ... Our thumbnailing even strips EXIF copyright information. :(

Well, that at least could probably be fixed easily enough. Of course, even if they just hotlink the thumbnail with no comment, you can still look at the URL and track down its contributors, if you know how. Rather like how in Wikipedia articles, the images are used without clear attribution, and you have to know to click the image to track down the contributors . . . although there's a difference of degree here, it's true.

I strongly suspect that an overwhelming majority of the hotlinkers are not following the rules. If we can't get that turned around then we *should* disable hotlinking because leaving it there is an attractive nuisance which is both disrespectful to our contributors, and potentially bad for PR (imagine if some Wikimedia contributors start aggressively DMCA noticing hotlinkers who don't provide attribution).

Which is kind of like saying that if an overwhelming majority of third-party reusers of Wikipedia dumps are not following the rules, we should stop providing the dumps. Wikimedia needs to stay within what is legal and moral itself. It doesn't need to punish legitimate reusers because of a majority's illegal or immoral actions. If contributors want to DMCA hotlinkers who don't provide attribution, that sounds like a good idea to me.

But ... I don't think we should limit hotlinking.

Rather, I think we should encourage it... but we should encourage doing it right by promoting extensions that provide links back to the image pages, and by having mediawiki provide example HTML copy and past which does the right thing.

Now there's a good idea. Every other image-upload software package does that. Let me see if I can't code that up right now.

On Mon, Jul 28, 2008 at 12:20 PM, Gregory Maxwell gmaxwell@gmail.com wrote:

One click vs.. I don't agree that it's at all comparable

Well, in neither case is it obvious that the credits are even available at all, and I think that's the important similarity. But that's neither here nor there. If we're okay with the image merely linking to the image page on Wikipedia, it should be okay for hotlinks too.

(Why are the links not something simple anyway, again, like http://commons.wikimedia.org/thumb/Image:ImageName.png/800x600? That would *almost* allow replacing "thumb" with "wiki" to get the image page . . . there's no reason to have it retrieve directly from the filesystem, when everything is cached by Squid anyway.)

But see in the hotlinking case we're an active participant. It doesn't continue without our help.

You mean, it doesn't continue if we decide to actively (possibly at a hit to performance) go out of our way to try to filter out those images. And even then it still probably continues, just people have to reupload it somewhere -- without *any* ability for the interested viewer to track where it's from, tech-savvy or not.

Besides, as I pointed out.. there are clear actions which we can take to mitigate the harm: good linking instructions, offering extensions to popular blogging platforms, and preserving/filling out image metadata.

That I agree with. :)

Cool. 'nuff. You might want to hunt down mangus little JS example. I can't find it at the moment, but the user interface was pretty reasonable as I recall.

Is there any problem with something that just looks like Flickr? Actually, I don't see any interface for this on Flickr -- say ImageShack, then?

Where should the instructions be put, right under the image?

Do we have agreement that we actually want this? It's rather a big leap in the opposite direction from the OP. :)

On Mon, Jul 28, 2008 at 2:19 PM, Simetrical Simetrical+wikilist@gmail.com wrote:

On Mon, Jul 28, 2008 at 12:20 PM, Gregory Maxwell gmaxwell@gmail.com wrote:

...

One click vs.. I don't agree that it's at all comparable

Well, in neither case is it obvious that the credits are even available at all, and I think that's the important similarity. But that's neither here nor there. If we're okay with the image merely linking to the image page on Wikipedia, it should be okay for hotlinks too.

I'm not sure we can have a reasonable discussion on this point. So I'll drop it for now.

[snip]

You mean, it doesn't continue if we decide to actively (possibly at a hit to performance) go out of our way to try to filter out those images. And even then it still probably continues, just people have to reupload it somewhere -- without *any* ability for the interested viewer to track where it's from, tech-savvy or not.

Possibly at a hit to performance. Come now. The CPU on the squids are loafing, applying a couple of regexpes to allow some specific sites would not be a hit worth mentioning. (and if it is, it would just be something else to fix in squid...)

So don't feed me that crap. ;) Besides, as I pointed out.. I'd rather not disable it.

...

Cool. 'nuff. You might want to hunt down mangus little JS example. I can't find it at the moment, but the user interface was pretty reasonable as I recall.

Is there any problem with something that just looks like Flickr? Actually, I don't see any interface for this on Flickr -- say ImageShack, then?

Where should the instructions be put, right under the image?

Do we have agreement that we actually want this? It's rather a big leap in the opposite direction from the OP. :)

This was the JS hack I was talking about: http://commons.wikimedia.org/wiki/Image:African_penguins.jpg?withJS=MediaWik... (helps to spell Magnus' name right!)

Looks like it's not working quite right anymore. The 'use on your webpage' link popped up a box with HTML that you could copy and paste. I think the flickr style size links on top is preferable ... less likely to get missed.

On Mon, Jul 28, 2008 at 2:37 PM, Simetrical Simetrical+wikilist@gmail.com wrote:

On Mon, Jul 28, 2008 at 2:32 PM, Gregory Maxwell gmaxwell@gmail.com wrote:

...

I think the flickr style size links on top is preferable ... less likely to get missed.

I don't actually see anywhere to copy and paste HTML to hotlink images from Flickr. I only see a download link, and that only after clicking a tiny and easily-missed "All Sizes" link just above the image. Does the HTML only appear on some images?

If we're talking about likelihood of getting missed, ImageShack's approach sure seems to be pretty noticeable:

http://img519.imageshack.us/my.php?image=picture1sp3.png

But that uses a lot of screen real estate.

I'd prefer something at the top which, when clicked, brings up a page with stuff like your imageshack example.

On Mon, Jul 28, 2008 at 2:46 PM, Andrew Gray shimgray@gmail.com wrote:

It appears underneath the image when you've gone through to "all sizes". It's there for all images with CC-* permissions, and - I think

• authors can set it to be visible even for all-rights-reserved images

in their own preferences.

http://www.flickr.com/photos/shimgray/2710860588/ is the last image I uploaded (don't worry, it wasn't all mine...) and "all sizes" will give you: http://www.flickr.com/photos/shimgray/2710860588/sizes/m/ a download link, then the image, then underneath that the raw image URL and a set of preformed HTML.

This is what I see:

http://www.twcenter.net/~simetrical/Flickr_no_controls.png

2008/7/29 Simetrical Simetrical+wikilist@gmail.com:

Is there any problem with something that just looks like Flickr? Actually, I don't see any interface for this on Flickr -- say ImageShack, then?

Where should the instructions be put, right under the image?

Do we have agreement that we actually want this? It's rather a big leap in the opposite direction from the OP. :)

YES! Great idea. Please give it API stuff too.

Brianna

2008/7/29 mboverload mboverloadlister@gmail.com:

I think we need to make it clear what can and can not be uploaded and promptly delete anything that doesn't fit those guidelines. This eliminates almost all "image hosting". The rest is so small as to be meaningless.

There's already Commons regulars taking care of this there; people who've mistaken Commons for Flickr.

- d.

2008/7/28 Simetrical Simetrical+wikilist@gmail.com:

Well, that at least could probably be fixed easily enough. Of course, even if they just hotlink the thumbnail with no comment, you can still look at the URL and track down its contributors, if you know how.

It's doable, but it's a real hassle and requires you to know the Secret Naming Conventions to handle it. I certainly wouldn't expect most experienced mediawiki users to know how to do it without experimentation, much less the average user trying to figure out the source of an image...

There has been some discussions with people within newspapers, archives and libraries in Norway to create some sort of common metadata standard so it can be harvested by Mediawiki. Especially to build "smart" reference tags, and to track news article that are on the move to a permanent newspaper archive. It seems like the "standard" will be something like Dublin Core, and some fall back solutions...

John (one of the persons..)

2008/7/28 Gregory Maxwell gmaxwell@gmail.com:

...

Rather, I think we should encourage it... but we should encourage doing it right by promoting extensions that provide links back to the image pages, and by having mediawiki provide example HTML copy and past which does the right thing.

A lot of copy & pasting does include links to the image description page on our servers, which I agree should be good enough for third parties if it's good enough for us. But I like the idea of easy HTML snippets for embedding.

What I'd also love to see is a reference implementation in MediaWiki of what we consider the ideal use scenario of free media from Commons in third party wikis: searching an image, embedding it, and making the metadata accessible. With the recently added foreign file repository support, any MediaWiki installation can now easily access images from Commons. It seems like the next logical step would be a cross-wiki search interface, perhaps similar to the one that Wikia has designed for its image search.

We could then encourage makers of other content management systems to implement a similar mechanism to add free media to content.

The best way to get people to do the right thing is to make it easier than doing the wrong thing. ;-) Fortunately, downloading and re-uploading an image into a CMS is more effort than having a direct web interface to acquire it, which should theoretically make it possible to pass along any metadata whenever the image is used. -- Erik Möller Deputy Director, Wikimedia Foundation

Support Free Knowledge: http://wikimediafoundation.org/wiki/Donate

---

foundation-l mailing list foundation-l@lists.wikimedia.org Unsubscribe: https://lists.wikimedia.org/mailman/listinfo/foundation-l