DOI best practices

DOI best practices

DOIs are the source of a large number of misconceptions that are unfortunately widespread and lead to complicated and unjustified mechanisms being put in place.

Through this page we try to bring our vision of good practices around DOIs.

General

Why do we make DOIs?

The sole purpose of a DOI is to be able to cite a digital object (a dataset, a website, a PDF file,…) in a permanent manner.

The DOI is in fact just a redirection to a web page called a landing page. What makes the DOI so interesting is that the pointed landing page can evolve over time.

Why not use a URL?

A URL is a good way to designate a digital object. However, they have two limitations that make them not good candidates for citation:
1- They are very fragile. Indeed, it is enough that the site changes its name, or that a page is transferred to another web site and the URL that you had communicated does not answer anymore
2- They are often long which is not really convenient for a citation in an article

Obligations

What are the obligations when setting up a DOI ?

The only obligation is that there should always be a relevant landing page for the person wanting to consult the DOI.

Example

A scientific project wishes to cite its activity. It requests that a DOI be created. As the project has a website, it is decided that the DOI landing page is the project website. After a few years, the project is terminated and the website is closed. The DOI maintainer will have to set up a web page (on another website) indicating for example « the project you are looking for is now closed. However, you can contact these people for more information… ».

This page must be declared as the new landing page for the DOI.

The main misconception about DOIs: versioning

The most common idea around DOIs is for datasets. People often think that once a DOI is set up on a dataset then this dataset should not be modified: no data can be added, existing data cannot be corrected.

The origin of this idea is the desire to ensure the reproducibility of science, which is of course a good approach. However, there are several obstacles that make this a bad idea:

  • Firstly, it is not necessarily possible, for reasons of storage volume, to keep all versions of a dataset
  • Secondly, people who set up these mechanisms often do so by regularly extracting data in the form of files. The issue is that, after a certain period of time, this files may no longer be readable because the managers do not make them evolve after they have been created while their data system is evolving.
  • Thirdly, depending on the case, this leads to the management of several tens, hundreds or even thousands of DOIs for a dataset. A number that is far too large to ensure proper management.

If accurate dataset versioning is needed and possible, it is primarily a problem – a complex one – for the data management system and not for DOIs. This will be discussed in the following paragraphs.

Good practice around DOIs.

The Research Data Alliance has summarised in a very clear document the good practices
https://rd-alliance.org/system/files/documents/RDA-DC-Recommendations_151020.pdf

At the French level, this document has been adapted and is available here (english version): https://permalink.sedoo.fr/doi_best_practices

The SEDOO implements most of the recommendations of the French document as you can see here: https://www.sedoo.fr/doi/

Among the main recommendations, here are some important points:

1- The DOI is set up by the organization managing the data
2- A DOI must have a corresponding metadata record
3- The DOI standard defines the notion of fragment. A fragment is an additional string that is appended to the DOI after a #. These fragments indicate which part of the dataset we wish to cite.

For example, if we have the DOI https://doi.org/10.XXXX/12345 which cites a dataset, we can imagine fragments #v1 or #v2 which specify the version of this dataset, the fragment #p2014-2019 which could indicate a particular temporal extension.

Thus the different DOIs would be available and would lead to the same landing page:

  • https://doi.org/10.XXXX/12345
  • https://doi.org/10.XXXX/12345#v1
  • https://doi.org/10.XXXX/12345#v2
  • https://doi.org/10.XXXX/12345#p2014-2019

The use of fragments makes it possible to greatly reduce the number of DOIs to be managed.

4- This fragment can be interpreted by the landing page in order to present to the user exactly the data mentioned. This is for example what we are doing in the future BGI citation system. This can also be seen – very modestly – with the following doi: http://dx.doi.org/10.6096/SSS-LEGOS-GRID-ATL#v2.0

On the metadata sheet, the download tab only allows you to download the latest version of the dataset. However, the « Versions » section displays the different versions of the dataset and the modifications they correspond to. The one corresponding to the transmitted fragment (v2) is highlighted by a different colour.

Rechercher