When thinking about "Open Data" (possibly due to nothing more than my own ignorance) I am utterly confused as to what difference an open licence legally makes for structured data.

"Open Data" for me is a bit like saying "Free Air".

From discussions with other (fairly knowledgeable) people on the matter (for example, this nice summary), it seems I'm not alone in my confusion. Before I ask the concrete question (which will be: "why should I care how a structured dataset is licenced?") I feel I need to cover my own potentially flawed understanding of the matter (so sorry in advance for the long question).

So let's say we have a dataset—let's say the IMDb dump. I am very interested to use the IMDb dump. The IMDb dump is publicly available for download from the Web. Great! ... but it doesn't have an open licence. Hmm.

But wait ... what legalities actually stop me from exploiting the factual content of the IMDb dump regardless of licencing or what the providers of the dump have to say?

As far as I can tell, legally speaking, assuming the dataset is comprised of factual content and that it's the factual content I am interested in using (e.g., the movie Dune was 137 minutes long) the only thing I have to avoid is duplicating the dataset in its current expression elsewhere or duplicating the expression inherent in natural language or images (e.g., the prose of an IMDb review or an IMDb image).

As I understand it ...

Licences themselves are a mechanism to outline the terms under which the licensor agrees not to sue the licensee. Independent of the terms of the licence, the licensor needs some legal basis on which to sue for a licence to have any purpose. The relevant laws are (presumably) Intellectual Property laws. Wikipedia gives a broad list of such laws (under CC-BY-SA) ...

enter image description here

You can't patent the first name of the Queen of England or the run-length of the movie Dune. So let's skip patenting. And trademarking.

A more relevant law would seem to be copyright. But at least in the US, and I imagine in many other jurisdictions, you can't copyright the first name of the Queen of England or the run-length of the movie Dune either. Because copyright does not apply to facts. Here's a quote from the US copyright FAQ:

"Copyright does not protect facts, ideas, systems, or methods of operation, although it may protect the way these things are expressed."

In a nutshell, copyright thus applies to expression, not to content. So okay, copyright can prevent you from copying the expression of the data verbatim (e.g., redistribute the exact PDF or the exact Excel sheet or idiosyncracies of the HTML page or an image or the exact text of some prose) but does not protect the factual content itself. If the data are mapped into a generic format like CSV, JSON, RDF, etc., copyright seems to have no jurisdiction over data.

So if not copyright, what else?

None of the other IP laws listed above seem relevant, except for the one you can see I clicked on from the image above: database rights. It seems that database rights vary widely from jurisdiction to jurisdiction. US database rights consider factual content as out of scope. Only the EU has database rights that should be of concern, here quoting from the Wikipedia article:

"While copyright protects the creativity of an author, database rights specifically protect the "qualitatively and/or quantitatively [a] substantial investment in either the obtaining, verification or presentation of the contents"

The EU database rights seem a little more "strong" than copyright in that they explicitly apply to databases storing factual content, but other than protecting the database as a complete curated package (i.e., the contextual/physical schema and its extension as tables), it does not protect the individual items of content in the database. You cannot just put a bunch of content into a database and then say that those contents are protected under database rights: precedent for people annoyingly trying to do this suggests that that's not how the law works.

Aside from IP, another possible legal facet is the Terms of Service/Acceptable Usage document you must implicitly/explicitly agree to when visiting the site to get the data, but other than banning the user from returning to a site or stopping them from accessing a service, I see no legal grounds on which a user can be sued for using a publicly downloadable dataset (again, other than infringing on the copyright of its expression).

So, assuming I'm not going to re-distribute the dataset with the same "expression" or re-use any creative elements likes images or prose writing ... assuming I'm only interested in using the factual content, my question then is:

Why should I care how a structured dataset is licenced?

The only meaningful answer I've come across so far is "so you know someone isn't going to try and sue you", which though a strong practical argument, I don't find conceptually compelling.

(I would be particularly interested in precedents where a party was successfully sued for using the factual content of a structured dataset that was not openly licenced.)

  • 1
    I hope that @RufusPollock will eventually provide an answer as well. In the meantime, have a look at his quite informative Guide to Open Data Licensing. – Patrick Hoefler Jul 4 '13 at 19:01
  • From a scan through, that's a very informative link and a useful reference, thanks! The document argues that providing an open licence for data is important for "clarity", which is fair enough. I wonder though ... would it not provide more clarity to say that factual content need not be licenced at all (rather than getting into the higgledy-piggledy of trying to interpret the implications of specific copyright licences for data). – badroit Jul 4 '13 at 19:27
  • Speaking from a European perspective, "clarity" is not really an issue, since European databases are quite clearly protected under sui generis intellectual property law (as explained in your linked Wikipedia article). For this reason alone, Open Data licenses are essential. If they bring clarity to databases that actually wouldn't need to be licensed, that's just icing on the cake. – Patrick Hoefler Jul 4 '13 at 19:52

IANAL but I believe most open licences are exclusively designed to provide reuse rights that were taken away by copyright (e.g. CC). Some licences also give rights that were taken away by the 'database right' (e.g. ODbL, OGL). (I don't have much knowledge about click-wrap agreements.)

I'm not sure that you can say that most data is factual, and therefore copyright does not apply. Track names for CDs are creative works. Weather readings and locations of items on a map have been defended vigorously against copyright (e.g. before database right came along in '96), although you'd have thought that because they are collected by strict rules, they'd fail the 'creative judgement' test. That the case of the football fixture list had to be referred all the way up to the European Court suggests it is not an easy judgement and that you have to be extremely careful. The telephone directory case also seems hard-won.

'Database right' does not apply to an individual item of data, but the threshold is somewhat lower than the entire database. As soon as you extract a substantial quantity from the database you become liable. So anything more than one item might invite a letter from the lawyers.

Since reusing data without a licence is such a mine-field, anyone who doesn't have a massive legal fund might as well regard it as a no-go area. That's why there is a big push by governments to provide open licenses for their data, to encourage transparency and reuse of these valuable national assets.

It is important to note that you can use data for your own internal purposes, whatever the licence. It's only if you want to republish the data or derived data that you need to study the licence.


I'll address your last statement first. The process of methodically copying content from a website is known as "scraping". There has been at least one notable "scraping" lawsuit, Cvent vs Eventbrite (brief summary & some relevant court transcripts). Cvent sued Eventbrite for scraping data from their website, citing the anti-scraping language in its user agreement. Interestingly, the court said "the Terms of Use for Cvent's website are not displayed on the website in any way in which a reasonable user could be expected to notice them." Look at the above links for further details.

The above case highlights the distinction between so-called "clickwrap" (or click-through) and "browse-wrap" terms and conditions. As the names suggest, in the former instance a user is presented with some terms to which they must agree before being permitted to access the website. The latter refers to terms and conditions which are posted on the website and are accessible via a link (often in the footer, for example). This article describes how clickwrap agreements are generally held to be more enforcable, as you might expect, and cites a few examples.

I will probably struggle to answer your other questions to your satisfaction, but this article includes a few interesting comments on the legalities of scraping with regard to copyright (in the US). The relevant points seem to be:

  • the enforcability of terms of use is a largely untested gray area
  • although the organisation/formatting/arrangement of facts may be copyrighted, the facts themselves may not be, thus merely stating copyright on a website may not protect the content

Essentially, it's the wild west out there. On the one hand you have websites that in many cases might be overstating their ownership of content/data, and on the other hand you have those who scrape data because it's easy and they almost certainly won't get caught not to mention successfully prosecuted.

  • +1, thanks for the discussion! It partly covers my question: in particular, it helps confirm my own suspicions that Terms of Use on websites are far from the legally-binding contracts many people (sometimes) view them to be. I would say though that my question extends beyond scraping (which I would view as parsing/scraping content from individual HTML pages). For example, IMDb make their entire dumps available in structured formats under conditions that only seem to be legally enforceable in some small parts: imdb.com/interfaces. – badroit Jul 3 '13 at 15:47

It might not be "conceptually compelling," but what the laws says is just one dimension of this. The business judgement comes down to "what is the risk/reward/cost profile?" Some companies settle with patent trolls because it's cheaper than fighting. Others taken a principled approach and will litigate at all cost if they believe they are in the right.

At this point you've publicly disclosed that you've read their license in detail and plan to do things that it prohibits because you believe it's unenforceable in court. How much are you willing to pay to validate that belief?

If all you care about is the basic factual data, why not just use the Freebase movie data which is CC-BY licensed?

  • I understand this argument (being sued costs money, even if you're in the right), but again, unlike the patent trolls, I have yet to see a precedent for copyright of factual data having been defended by the provider. So what's the big deal? :) Instead of advocating open licences, why don't the community (also) advocate that data doesn't even need licencing? My counter-question again remains: if all I care about is the factual data, why should I care how the data are licenced? (The question is not specific to movies; just an example.) – badroit Jul 4 '13 at 17:21

Besides the large and complex interpretative aspects of copyright and all other IP variants, an open data license, such as the Open Government License in the UK plays a social role as well.

As you mentioned, an open license could be construed as just a way of telling you 'we wont sue you'but actually, that message is important, it gives confidence on re-use, it makes a covenant between holder and user that it is ok to re-use (changing behaviour towards openness), it links the re-use with any other domestic or international licensing framework and hopefully harmonises it and it provides security, when re-using to create services and apps, that will probably be required by investors, cautious perhaps of getting bitten later on by an unexpected legal case.

An Open Data License is an explicit admission that it is ok to re-use, a public commitment by the IP owner which as behaviour change goes, I think plays a big part on making more data open.

It also does away with all the uncertainty malarkey so you can get on with doing great things with that data.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.