I see two problems here: First, while this may be legal in the European Union, it is far from clear for all jurisdictions around the world. See for example Copyright aspects of hyperlinking and framing - Wikipedia. I suppose we should not use such deep links in our data, on the basis of “whiter than white”.
Secondly, apart from the legal considerations, it is clearly an unfriendly act towards a site owner who does not want his content deeplinked. I’d prefer it if OSM didn’t take such an unfriendly stance.
The extraction of any text, photos, images, graphics for use outside of this website, including in the frame of an external site or in the form of deep links, is strictly prohibited without permission.
I assume it means the author doesn’t want image links to be shared, but a link to the home page shouldn’t be a problem. That doesn’t solve everything, of course, because we may or may not still have a link to an image somewhere.
Is there a list of countries with laws that say links can somehow be protected? Because then we can limit our scope of this topic to just those countries.
P.s. the existence of the image=* tag in OSM is also a problem, because mappers use it to link to images with all sorts of copyright protections. Anyone using that data is likely to breach some copyright protections.
Indeed. The wiki page on photo linking cheerfully states “Creating a link between an object and its picture in OSM is easy.” and lists image= as an option for "linking to any image" without highlighting that it is a permissions nightmare compared to the other options in the table.
I’m not a lawyer, but that’s likely to depend on jurisdiction. I think it’s true in the EU via an ECJ judgement, but I’m not sure if that still applies in England and Wales.
That said, linking via an image tag to a site that does not want to be hotlinked to is a Bad Idea for other reasons - not least that the image that you actually get may not be the image you want (that link is just to text, but if you are of a pearl-clutchy disposition you may not want to click the links there).
The cases in the Wikipedia page are about the protection of works that were embedded in other’s websites. This site owner, on the other hand, refers to the protection of his urls. As far as I know there is no such thing as protection of urls.
Are we sure the author really is opposed to any links to his website which point to somewhere other than the main landing page? I would have assumed he was only concerned about deep links which would lose the context of the site, such as direct links to images. The kind of links we have in OSM aren’t really an “extraction […] for use outside of this website”.
In general, I would have assumed it’s quite rare that website owners would be opposed to inbound links like that. But if that is really a site owner’s point of view, I would likely be inclined to honor it. At least in the case of third-party websites such as this one.
While it’s not directly related to the question of this thread: website=* is intended for the official website for a feature. I don’t think that private website is the official website of the features it is linked from, so I would consider the use of the website=* key incorrect. If we want to keep the links, it would be possible to use url=* instead.
Even if image=* isn’t an immediate legal problem for us, it is riskier for data consumers from a legal perspective, moreso if a data consumer does the obvious thing and embeds the image instead of linking out to it. Either way, it’s a security and performance problem for end users. And we’ve stripped any metadata or context from the image, making it all but impractical for a data consumer to insulate themselves with disclaimers or filters. So I personally put this key in the category of nice ideas that don’t work.
(Sometimes it is necessary to cite a photo as proof that something is mappable, but that’s what source=* is for.)
What concerns me is that the context is lost once we put a (deep) link into our database. OSM users need to be able to trust that OSM data can be used freely (within the licence). No data user will be aware that - in contrast to other links in OSM - these specific deep links are not meant to be used for “extraction of any text, photos, images, graphics for use outside of this website”, or that their use as such might even be illegal in some jurisdiction.
Hmm it isn’t as if the situation is any different for any other links or references.
We (as in OSM) do not make any representations or statements on third party content not included, but just referenced by/in OSM data, it could have literally any terms of use or licence and (because we don’t include any meta data to that respect) the data consumer needs to determine this themselves.
That’s true as long as this is just a link. But with any key named image=*, the naïve and apparently common approach is to embed the image on the map or in a popup of some sort. Depending on the size of the original image, which the application doesn’t know ahead of time, it could be a thumbnail or the image at nearly full resolution. None of this puts us in any legal risk, but it feels like a bit of a trap for data consumers compared to other keys like website=*.
I don’t quite see what that changes wrt the data consumer needs to determine this themselves. That they don’t do that is, well their problem, and they’ll learn latest when they get a copyright claim against themselves.
Sure, that is technically true, but also somewhat beside the point of this thread, which is about best practices. Those of us who care about promoting direct and indirect usage of OSM and upholding the project’s reputation would also care about making it easier to use the data correctly.
For example, we could accompany any image=* with some licensing metadata in the form of an image:license=* tag.[1] Even better if that tag contains machine-readable SPDX or Wikidata identifiers. In the absence of licensing metadata, a data consumer would know to tread more carefully with the image, only treating it as a link. All this needs to be documented clearly so that a naïve data consumer won’t do the obvious thing and potentially hurt OSM’s reputation.
However, this still doesn’t address the security, privacy, performance, and reliability issues inherent in hotlinking (embedding) full images. That’s why overall I would prefer data consumers to discover images through linked open data (via wikidata=* etc.) and let those other projects worry about these issues.
there’re also trademarks in our map data, which you might not use “freely”.
I agree that we should encourage linking only to images which are available under a free license, but it’s impossible to guarantee for it.
