You don't have to open source or share the code you use when working with this data

IANAL but this feels a bit vague to me

Extraction could include querying a copy of the database, working with an API, etc.

Substantial quality has a reasonable interpretation, but its not clear what "substantial in quality" might mean

The word Substantial is significant through this licence. You'll notice that its used in multiple definitions. E.g.:

• Extraction refers to Substantial parts of the dataset
• a Produced Work is the result of the whole, or a Substantial part of the dataset
• Re-utilisation application to using or or a Substantial part of the dataset

So some uses of the data only trigger clauses of this licence if there is Substantial use of the dataset. However if you make lots of small uses, e.g. repeated queries or extraction, then these are considered to add up to a Substantial use.

In short, Substantial means accessing a large quantity of the data either in a single operation, or through a number of smaller operations.

Unfortunately there's no real definition of Substantial in terms of how much of the dataset you might be using. E.g. a percentage of its contents. This would be tricky to define in general, but does mean there is some ambiguity around interpretation: at what point to I step from using small amounts of data to using a Substantial portion? When multiple operations can stack up to a "substantial" use, then this gets even trickier.

This section clarifies that there are some legal rights that you may have, e.g. right to use information for research purposes that are outside the scope of this licence. This section also explicitly makes clear that "insubstantial" uses of the data aren't covered by the licence.

This clause spells out what Share Alike means. Basically either provide the derived database directly, or the means for someone to recreate your version of the data.

If you create something using a derived database, and you only use that within your organisation, then you don't have to share your derived database.

This makes it explicit that if your use of data under the ODbL is deemed to involve the creation of a Collective Database then the Share Alike provision doesn't apply. Although you will still have to provide notices to attribute your source

If you publish a work that is based on a derived database, then you need to also share the derived database.

No compatible licenses have been declared by Open Data Commons.

The later clause, 4.4e states that Licensors can authorise someone to determine if another licence is compatible. Although its not clear that anyone has done this.

As the publisher of the data (the Licensor) you can choose to publish your data under alternative arrangements. ODbL chooses to make it explicit that this right is reserved, although technically this is unnecessary as the final clause here makes it clear that all other rights are reserved

This gives explicit permission that you can use ODbL licensed data in commercial applications.

This is an important distinction between the ODbL and, say, the Creative Commons licences which rely on copyright law.

The thing you have made with the data. This might be an image, an artwork, an animation, a t-shirt, etc.

This is the second key concept required to help interpret the ODbL. The first being "Collective Database".

Again, the definition identifies a set of operations might might result in a derived database, but there's still some ambiguity. This result in frequent questions to the OSM community.

When you "convey" a database according to this licence, you'll need to do certain things, e.g. attribute. This definition is attempting to clarify what "convey" means.

This is the first concept that is key to understanding how to comply with the ODbL. Unfortunately a "collection of independent databases" can be interpreted in several different ways.

For example: a set of separate data files or database exports that are packaged together, e.g. in a zip file, meets this definition. We might collect works together like this to help to distribute data.

The text also notes that the database is in "unmodified form" but doesn't elaborate on what that means. It seems reasonable to assume that creating a subset of a database, e.g. as the result of querying the data, is not modification. But changing or correcting values is modification.

The OpenStreetMap community have a whole wiki page that elaborates on what a "Collective Database" means for their data.

The examples on that page indicate that, for the OSM community at least, a Collective Database also includes a dataset that merges OSM data with other data, so long as the new data either completely replaces any similar data available in OSM. Or is an entirely new set of fields. In this case, you don't have to publish your data under a share-a-like licence.

This example might not what people expect from a share-a-like licence when applying it to their data, or when looking to use a dataset like OSM. They may actually have more freedom than they expect.

Some people might consider this to be a Derived Database (see below). If you're apply the ODbL to your data, you should probably clarify your expectations.

Another example would be third-party rights that apply to some (or all) of the data in your dataset. This might be the case if your dataset contains data from other sources. Or has been validated against them.

Basically if you have multiple licences that apply to the images, text or other media in your dataset then include a column or field in the dataset that indicates which licence should be used. Or, failing that, some way to identify which licence applies to a specific subset of the content.

If your dataset includes text or images along with the data then you need to combine the ODbL with another licence. For example, if you have created a collection of images and annotations to be used as training data for machine-learning applications, then you can apply the ODbL to the annotations. But should also indicate which licence applies to the images. If you instead use a Creative Commons licence then you can apply a single licence to the dataset and the images.

At the time that the Open Database License was created the Creative Commons licences were only applicable to works that were covered by copyright. As the licence notes, the protections for databases vary across different jurisdictions. A licence designed for databases needs to cover these other types of rights. The ODbL was the first licence to address this. However since then the Creative Commons 4.0 suite of licences have been released. These now address database rights so can be used to openly licence databases.