5 things you may not know about DOIs or why there is more to DOIs than meets the eye
Note: This is adapted from an internal talk I gave at my place of work. It is written for generalists who want to have an idea about DOIs and related services and hence does not go into specific details (e.g. how to actually deposit metadata, fields used etc).
The thesis here is that if you are an academic librarian you could probably do well to learn more about DOIs. As one of the central pillars of Scholarly infrastructure, DOIs are probably more important than you may realize.
Whether you are
a) an information literacy librarian wanting to educate in depth about preprints (do they have dois? how do they relate to published version of record articles?) or retractions (how are retractions captured? How do you check for them?)
b) a liason/subject librarian trying to answer a question about text mining of journal articles or about coverage of tools like Lens.org vs Scopus (Lens draws from Crossref as a major source, how complete is the metadata?)
c) a electronic resource librarian trying to understand the role of DOIs in linking technologies like OpenURL, GetFTR etc
d) Institutional repository manager wanting to understand how to get data citations from your respository to be captured properly (See MLibrary's Susan Borda try to figure out data citations flow through the publishing process)
knowing more about how DOIs (and PIDs in general) work in conjunction with other pieces of knowledge may give you a fuller understanding of the topic at hand.
If all this sounds unlikely to you, read on.
Structure of a DOI
DOIs or Digital Object Identifers have become ubiquitous in our Scholarly ecosystem to the point that I find most researchers today are at least vaguely familar with the idea. You can possibly blame Sci-hub for this, but today many other tools and systems are comfortable with asking for DOIs without feeling the need to explictly explain what they are.
Similarly librarians even those not in the Scholarly Communication know of DOIs. Most of us know DOIs are meant as a unique persistant identifer for articles but having a DOI has no bearing on the quality of the article.
Still knowing of DOIs is not the same of actually having good or detailed understanding.
Until fairly recently most libraries and librarians were not publishers and did not have direct experience with minting of DOIs and as such I suspect many librarians still have a somewhat surface understanding of what DOIs are.
DOI for more reliable linking
For most , I suspect we see DOIs mainly as a type of Persistent Identifer that we can use to link to the Scholarly content (typicallyjournal article) of interest.
Resolving DOIs
We appreciate that using the DOI to lookup the URL is more reliable than linking to the URL directly.
The idea here is whenever the location of the scholarly content on the web changes, the content owner is obliged to update the link the DOI resolves to. As such as long as we point to the DOI, we will be sent to the correct updated URL, which reduces link rot.
The DOI generally resolves to one URL, but multiple resolution where the user is sent to a landing page with multiple options while rare does exist.
Updating where the DOI resolves to can be due to a change in the internal structure of the publisher website but can also be due to the content owner selling the journal to another publisher
It is often assumed that each unique content is registered with one and only one DOI and while this is usually true, there are exceptions (and this isn't even relating to preprints/postprint confusion). Crossref's own analysis in 2020 detected 0.8% duplicates. Various reasons for this. For example if a publisher gains control of an existing journal, they should check to ensure control of the corresponding dois are transfered to them as well and to update the links according, however some publishers may just mint their own DOIs leading to duplication.
In a way this concept of DOI is somewhat similar to an idea of a shortener URL like bit.ly which redirects you to the right URL. (Side note there are short DOIs!).
This concept isn't wrong and this was indeed one of the main reasons DOIs began to be used in the Scholarly system (see quote below),
As the linking hub for scholarly content, it’s our job to tame URLs and put in their place something better. Why? Most URLs suffer from link rot and can be created, deleted or changed at any time. And that’s a problem if you’re trying to cite them.Thus the Crossref DOI was born: an Identifier which is Persistent, which means that it’s designed to live forever (or, as Geoff Bilder rather more prosaically puts it, as long as we do), and also Resolvable, which means that you can click on it.
However as you will see there is more to DOIs than meets the eye.
DOIs (and PIDs) in general today are used for more than just resolving to content
For one thing, to even do the simple function of being a Persistent ID (or PID), metadata needs to be maintained for each DOI. At the very least you would need the DOI and the URL to resolve to. Obviously the DOI would be also associated with the usual metadata describing the digital object, e.g. title, author etc.
But as you will see later the metadata associated with each DOI has expanded to cover multiple aspects about the object resulting in interesting and important services built around them such as supporting updates/retraction checks, Plagarism checks and finding citations.
The key point is DOIs (particularly Crossref and Datacite ones) can express relationships between DOIs and since DOIs can be used for a variey of objects beyond published journal articles, such as for datasets, preprints, peer reviews, grants and more, you get many interesting possibilities.
Some DOIs in crossref crosslinked by relationships
For example you could link a dataset with a Datacite DOI to a published Version of Record journal article with a Crossref DOI with the relationship isSupplementTo. Or you could link a preprint paper with a Crossref DOI with yet another Crossref DOI for the published Version of Record with the relationship isPreprintOf . Have a peer review report with a DOI? Connect it to the published paper the peer review is on with isReviewOf
Once you have such a network of research objects connected to each other, we can create networks fo scholarly objects, what Crossref calls a "article research nexus" that can be used to query to answer relevant questions.
And of course, DOIs are only one type of PID (persistent ID). DOIs can include other PIDs in their fields such as ORCID for authors and ROR for affiliations and by collecting relationships between such PIDs, this results in the PID graph.
Want more details? Read on.
Over the years, as I started to learn more about DOIs (in particularly Crossref DOIs), I was amazed on how much more there was to learn about DOIs. Below are some of the things that surprised me about DOIs.
1. DOIs are not issued by just Crossref, other DOI registration Agencies exist
2. DOIs are not just about links but also about metadata
3. Crossref DOIs have many services beyond just content registration for linking e.g. Similarity check, Crossmark
4. Crossref DOIs allows content registration for not just journal articles but also posted content (preprints, blogs) ,grants and more.
5. Crossref DOIs allows creation of relationships between DOIs e.g. preprint to VoR (Version of Record), Dataset to journal article, Peer review report to article etc
Not all these distinctions will be useful to you in your work as a librarian or researcher, but some may.
Also see Crossref's own mythbusting , though this is more publisher centric.
1. DOIs are not issued by just Crossref, other DOI registration Agencies exist
I think most people may have heard of Crossref when talking about DOIs and assume all DOIs are from Crossref.
In fact, there are 12 DOI registry agencies (RA) that are allowed to register or 'mint' DOIs. The most famous and popular ones used by Journal publishers is Crossref of course.But you may also encounter DOIs registered from Datacite and rarely other DOIs from others such as CNKI and Airiti (China content), KISTI (Korean), JaLC (Japan).
Some DOI registration agencies like Entertainment Identifier Registry (EIDR) , don't even cover scholarly content but are used to keep track of movies & TV shows! See list of coverage of all 12 RAs.
Ultimately all DOIs from the 12 RAs are governed under the system run by the DOI Foundation and each DOI is unique. If all you want is to resolve DOIs aka use a DOI to link/resolve to the content, you probably don't have worry that much about RAs and all you need to do is to type in https://doi.org/<DOI> and that will work regardless of which RA the DOI was registered under.
Crossref (and Datacite) as DOI RAs, will further assign range of DOIs by prefixes (see above) to content owners. So for example Wiley would be assigned DOis with prefix 10.1002. After that it is up to Wiley to decide how they want to assign DOIs to each piece of content in the suffix.
Technical side-note: In the early days of Crossref - suffixes were often "meaningful". For example they could be ISBNs or be the article accession number for the journal or just a combination of issn,vol,issue,page,date. Current best practice to create 'opaque suffixes' e.g. by randomising etc
Implication : So why care about which RA the DOI is registered with?
Firstly, while you can resolve DOIs from all RAs using the DOI resolver system, not all systems that claim to accept DOIs will work as some systems work only with DOIs from one RA , typically Crossref.
Take Unpaywall API, it claims to accept Crossref DOIs specifically, but what happens when you try a Datacite DOI or a KISTI DOI? Does it error out? or does it work and just resolve as per normal to the content? What if it needs to pull in metadata? (see later).
ThirdIron's Libkey.io - what DOIs does it work with?
The same applies for say Literature mapping tools that claim to accept DOIs as inputs etc. How about link resolvers or Third Iron's Libkey Link or Libkey.io? Do they only work with Crossref DOIs? If so can it find the paper on ResearchGate or Arxiv that uses a Datacite DOI?
The other thing to note is that DOIs minted by different DOI RAs have different metadata schemas. As you will see in #2 , when you mint or register a DOI, the content owner that does so deposits not just the link to the content but also different metadata. While Crossref and Datacite are two RAs that work closely (see later), they have totally different metadata schemas and hence different metadata fields that can be used.
As such it is important to note that by default when you query for metadata from a Crossref API for a Crossref DOI and do the same for a Datacite DOI via the Datacite search API, you will get totally different metadata schema outputs, that you may have to reconcile with mappings across the two different schemas if you have DOIs from both.
Also as you will see in #3, DOI RAs like Crossref and Datacite offers additional services to users of their respectively DOis that goes beyond minting a DOI and resolving the DOI to the link.
Again you will find if you try to use a service available only for Crossref DOis such as the antiplagarism "Similarity check" or check for retractions using Crossmark, it will fail if you use a non-Crossref DOI e.g. Datacite DOI
Lastly pricing and requirements for different RAs differ and if you support data repositories or any form of digital Scholarship and are asked about supporting PIDs you need to know the choices available. In our context, it usually means deciding between Crossref and Datacite DOIs.
Crossref and Datacite RAs distinguished.
Because this comes up a lot, I will include further details here.
Crossref and Datacite are two DOI RAs that you will likely encounter the most since they issue DOIs for scholarly related objects.
They collobrate heavily in everything from data citations to Event data API (see #3), still a DOI can only be minted under one RA.
Below are some notable organizations that issue DOIs of each type
Datacite DOI
Figshare
Zenodo
Dryad
ResearchGate
arXiv
Crossref DOI
Most major journal publishers
SSRN
medRxiv
OSF Preprints
bioRxiv
In general, the advice given is if you follow a "publishing" workflow you generally better off registering with Crossref and if you are a "Depositing" workflow you should register with Datacite, so you can see most data repositories like Zenodo, Figshare using Datacite and Publishers like Elsever using Crossref.
However this isn't definite if you look at the list above you may notice preprint servers most have chosen Crossref though arXiv uses Datacite.
It is possible for an organize to register with both and assign DOIs depending on the sitation. Some academic institutions for example both manage a Institutional Data Repository which issues Datacite DOIs as well as publish Open Access Journals which will use Crossref DOIs.
Another example is COS (Centre for Open Science), which generally uses Datacite DOIs for projects and other registrations (e.g. OSF preregistrations) but uses Crossref DOIs just for OSF preprints.
So why do organizations choose to complicate methods by minting DOIs belonging to two different DOI RAs? Simple, because they have different services and functionalities, see next two sections.
For the rest of the blog post, when I talk about DOIs, I refer by default to Crossref DOIs and sometimes Datacite.
2. DOIs are not just about links to the registered content but also about metadata
I used to have a simplistic idea of how DOIs worked and in this simple conception, the whole point of DOIs was to allow reliable linking. If you believe this, DOIs, such as those issued by Crossref would basically just need to keep track of two things.
Firstly the DOI itself and secondly the link whcih it resolves to , which will be updated when necessary.
In reality, getting a DOI minted, known commonly as Content Registration for Crossref DOIs is very much about content owners usually publishers depositing Metadata into Crossref.
As I noted above, a member of say Crossref, would be assigned a prefix of DOIs and they would choose what suffix they would want to use with it to assign to create a unique DOI for the content. However before they can complete content registration with the chosen DOI, they need to submit metadata to Crossref.
I can't do better than quoting Crossref here
Members send information called metadata to us. Metadata includes fields like dates, titles, authors, affiliations, funders, and online location. Each metadata record includes a persistent identifier called a digital object identifier (DOI) that stays with the work even if it moves websites. Though the DOI doesn’t change, its associated metadata is kept up-to-date by the owner of the record.
It is important to note that while a journal may begin life as under one publisher say Wiley with the associated DOI prefix, journals can switch publishers. If that happens, existing articles minted with existing DOIs will continue to use the DOI with prefixes of the original publisher (though newer articles after that will switch to the new DOI prefix). Assuming there is a change in where the content resides, existing DOIs would of course update the URLs where the DOI resolves.
Members maintain and update metadata long-term, telling us if content moves to a new website, and they include more information as time goes on. This means that there is a growing chance that content is found, cited, linked to, included in assessment, and used by other researchers.
So what metadata do Crossref and other RAs allow or mandate to be deposited? First thing to note is that different DOI RAs have different metadata schemas and requirements and this limits what additional thing you can do with the DOIs (see next section).
Here's the latest (as of time of writing) Crossref metadata schema 5.3 vs Datacite metadata schema 4.4. (Required, recommended, and optional elements for Crossref)
If you limit yourself to Crossref DOIs, you won't be surprised to see they accept standard metadata like title, author, journal etc.You may or may not know they also collect other PIDs like ORCID, ROR, funder ID, etc
But Crossref metadata is more than just the usual metadata and PIDs but also collects metadata that focus on relationships between different research objects. Some of the following metadata that Crossref accepts might suprise you, a limited list includes
References of content registered
Links to and from preprints
Full-text links (for text and data mining + similarity check - see later)
License URLs (for text and data mining)
Links to data and datasets, protocols and methods, clinical trial registry, trial number
Links to peer review report
Updates (such as corrections, retractions, updates, expressions of concern)
Funding data (conflict of interest)
Clinical trials
Technical note : Metadata in Crossref is not wholy from members, Crossref does a wide variety of things that automatically try to help enhance metadata submitted by members. As you will see later, they try to automatically link deposited references without DOI with DOIs for example.
In general, the mandatory requirements for most DOI RAs including Crossref are low (see list of mandatory fields for each Crossref content type), but in the case of Crossref, their participant reports helps us have a sense of how rich the metadata is being deposited into Crossref is by member.
Below shows a sample Participation report for a Crossref member - Cambridge University Press (CUP)
10 main items are checked, most of which are of great importance for discovery, research and editorial integrity and Reporting and assessment areas.
As you will see later, the history of Crossref has been an expansion in support of metadata being deposited into Crossref.
3. Crossref DOIs have many services beyond just content registration for linking e.g. Similarity check, Crossmark
While content registration is the primary function of Crossref which allows items to have DOIs minted, the fact that because Crossref collects a lot of metadata, it is able to provide many additional services to the Scholarly ommunity beyond just content registration and redirecting you to the webpage.
Some major notable services that work with Crossref DOIs includes
Ever wonder why Crossref is called Crossref? I suspect this refers to the fact that it is short for "Cross references" and all Crossref members are obligated to link references from their content to other Crossref member contents with DOIs. See below
Given that most style guides are slowly mandating the use of DOIs, this is becoming standard practice anyway, but Crossref offers services which are available to even non-Crossref members to submit references and quickly find and link to relevant Crossref DOIs.
Note that while there is an obligation of Crossref members to link to Crossref DOIs in references it is not mandatory to submit references in the metadata, see next service.
Crossref members can benefit from the citedby service. Essentially if you and other Crossref members deposit references with your metadata when minting dois, Crossref will be able to do the matching and calculate the cited count/total cites to your content.
This cited count is available to all (even non members). But as a Crossref member you can see not just the count but also the list of other Crossref works citing your content. One thing to keep in mind is until recently, content owners could keep the references deposited by them closed so nobody else could access them.
I've recounted many times on this blog how I4OC pushed for Crossref members (basically publishers) to make their references open (before that they were mostly deposited closed) and over the years reported on their growing success to the point that almost all major publishers in including Elsevier gave in on Dec 2020
We now live in an era of Open citations, where references for most academic content is open resulting in the rise of open citation indexes and services that rely on them.
In fact as of April 2022, Open Citation victory is so complete, that Crossref has voted to abolish the option of keeping deposited references closed. All references by default that are deposited are now open!
Somewhat basic but needs to be said. With all the metadata deposited by Crossref members, how does one access them?
This is where Crossref provides 7 diference features and options for anyone (most options are free except for Metadata Plus) to query and access the metadata.
For example, you can use standard web interfaces (there are two - Metadata Search, and Simple Text Query) but if you can code to some extent there are APIs you can use.
From standard RESTFUL APIs, to OAI PMH , there's even a OpenURL service (your link resolver is probably setup to work with it by augmenting or replacing metadata drawn from Crossref - example)
Sidenote : If you are a electronic resource librarian you definitely want to understand the idea of DOIs, understand why they don't solve the appropriate copy problem (even with multiple resolution) on their own and how they may work with your OpenURL link resolver.
Given that Crossref is one of the most if not most important source of open Scholarly metadata out there, with practically every major Scholarly service (e.g. Unpaywall) that uses such data relying on it, this is clearly an important service.
Side note on Text mining
It is important to note that among the metadata fields you can get from Crossref, you can also get "Text mining URLs" (this is one of the 10 items tracked in participant reports). The idea here is to make text mining easier by using DOI as a common way to a) Check licenses for text mining b) get URLs with direct access to the full-text across all Crossref members.
What is required for Crossref members who want to use this system is to deposit metadata for each item with licenses and URL that gives direct access to the item. Of course, each member will also need to figure out a way to authenicate access via the URL. In any case, if done properly a researcher can use one API, the Crossref API to text mine content across different content owners/publishers.
This is a relatively new and I suspect mostly unknown service jointly developed by Crossref and Datacite. Here's how Crossref describes it
When someone links their data online, or mentions research on a social media site, we capture that event and make it available for anyone to use in their own way. We provide the unprocessed data—you decide how to use it.
Above shows some sources captured.
While some of the events draw from sources like Twitter, Wikipedia, Reddit etc and have a distinct altmetric like feeling it is important to note that Events data allows you to find references between Crossref and Datacite DOIs, particularly in the area of data citation
Of course the difference between this service and slick commerce services like Plumx and Altmetric.com is that the Crossref Events data provides raw data about events via an API, so some work is needed to extract and massage the results.
Content registered by Crossref might change over time. Even if it is a published VoR (Version of record) of a journal article, the contents might be corrected or even retracted. Even if the content itself doesn't change, the metadata around it might. Articles may get additional metadata like information on ORCIDs, funding, peer review, license information etc.
How does the reader know if the article he has downloaded in PDF say 3 weeks ago is the latest? What if the paper got retracted in the meantime? This is where Crossref members who participate in Crossmark will embed Crossmark buttons on PDFs and on their journal platforms. In PDF format, if the user clicks on the Crossmark button they will be sent to a page providing the latest updates on the article.
As I have blogged in the past, reporting of retractions has come under a lot of scrunity in recent years. In particular, it was found that determining a paper had been retracted was not always straight forward. There are a myraid number of reasons for this, which led to a third party - Retraction Watch to start tracking retractions independently from Crossref's Crossmark system which was meant to do this. The Retraction Watch Database which now fuels retraction checks on various services from Third Iron's Libkey to Zotero,Endnote & Papers retraction checks.
Still over the years, I think improvements have started been made. One major response to the issue is the removal of fees for participation in Crossmark from Jan 2020. In the past Crossref members had to pay a seperate fee per DOI to use Crossmark which was clearly a disincentive. While this doesnt solve all the problems, at least financial incentives wont be in play in the decision of content owners to use Crossmark.
As a librarian, I was familar with the Turnitin service which is licensed by many institutions as a plagiarism checker of student assignments.
You might be interested to know Crossref members - typically Journal Publishers may also subscribe to the same technology (by the same company that offers Turnitin) called Similarity Check (formerly Crosscheck), powered by iThenticate to screen submitted manuscripts for plagiarism.
Crossref members that choose to use this service will have to pay a discounted fee (both annual and at a per document checked basis) but more importantly commit to having the full text of their DOI assigned content deposited in iThenticate database to be used for similarity checks.
in #2 we talked about how Crossref members can submit or deposit various types of metadata including Full-text links, which provide direct links to full-text content. One of the requirements for members to be on Similarity Check is to have at least 90% of such DOIs to have full-text links deposited into Crossref.
This is used by iThenticate to harvest the full-text content to be put into their database.
There's a lot of interest in "following the money", to see what Funder money is producing. Most authors do acknowledge Funders in different part of the full text often in non standard ways, so systematically extracting this information is not easy.
This is why Crossref is encouraging members to contribute metadata on the funder of works registered with a DOI. The problem of course is this Funder name must be controlled to ensure consistency and this is why Crossref maintains a Funder registry of all funders which you can search here.
Each Funder is assigned a DOI - for example here's Singapore National Research Foundation is assigned 10.13039/501100001381 and you can use this to search for all works that acknowledge this funder.
Ideally Crossref Members and Publishers should collect Funder information from authors during the manuscript submission system (via a form) and match them against the Funder registry and add the appropriate DOI to the metadata submitted.
More interestingly, while there are DOIs for funders, individual DOIs for grants have also started to gain attention.
4. Crossref DOIs allows content registration for not just journal articles but also posted content (preprints, blogs) ,grants and more
So what can you issue DOIs for? The obvious answer is journal articles and we have all seen journal articles on publisher platforms issued with DOIs.
But today the world is more complicated and that's not the only Scholarly content we can mint DOIs.
DOIs can be registered for book chapters, preprints, Conference proceedings , Dissertations
More surprisingly to some, Crossref DOis can also be assigned to
Dataset (this is more commonly seen as a datacite DOi in data reposit)
Preprints (example : https://doi.org/10.1101/2020.02.03.933226)
Peer reviews (example: https://doi.org/10.1162/qss_a_00146/v1/review3)
Grants (example : https://doi.org/10.37717/220020589)
There seems to be new developments and interest in assigning DOIs to grants and funding information see here and here. But my understanding of this area is hazy.
The fact that preprints can be issued DOIs is particularly tricky, given that preprints often do eventually get peer reviewed and get published as a VoR (Version of Record) so when do you issue a DOI and when do you not?
Crossref has you covered here
Essentially work "published" on preprint servers like SSRN, Arxiv will have their own DOI, and versions submitted for peer review up to the actual published version should have one other different DOI.
Given you are rarely likely to encounter #3 (pending publication), the more important point is in general Author Accepted Manuscript and Version of Record should have the same DOI, which can be surprising to some.
If you have signfiicant updates after final publication then you will need a new DOI.
5. Crossref DOIs allows creation of relationships between DOIs e.g. preprint to VoR (Version of Record), Dataset to journal article, Peer review report to article etc
In the above we saw that different versions of the same paper might have different DOIs. For example a preprint in SSRN would have a different DOI than one that is eventually published in the journal.
But can we somehow indicate the relationship between these two? Indeed we can via relationships between different versions of the same work.
Technical Note : Here I describe what Crossref calls typed relationships. There are other ways to express relationships using Crossref DOIs , including via references lists or using specific Crossref services like Crossmark, and Cited-by services. The main advantage of using typed relationships is that it works for creating relationships beyond just to Crossref DOIs but also other DOIs and identifers. Datacite in particularly uses the same type of linking which is handy for data citations
So for example, a DOI for a preprint could be linked to another DOI for the published version of record DOI which a isPreprintof relationship.
Specifically, preprints known as posted content in Crossref are obliged to link to link to the published version which this relationship. (There doesn't seem to be a requirement to link in the opposite direct from the Publisher version to the preprint version but there is a relationship hasPreprint available). But how do the content owners of the preprints know there is something to even link to? Crossref documentation notes
We will notify the member who deposited metadata for the posted content when we find a match between the title and first author of two publications, so that the potential relationship can be reviewed. The posted content publisher must then update the preprint metadata record by declaring the AM/VOR relationship. The notification is delivered by email to the technical contact on file
Besides the metadata, there are also some requirements on how the linking to the AM/VOR should be made on the landing page of the Preprint.
There has been studies on how often preprint do include link to the final published versions and the results show there is still room for improvement. Of course this alert only applies if both DOIs are registered under Crossref, but if say the preprint is registered under Datacite DOI (which may occur for say Arxiv) , it will obviously not get an alert if the published version of record is registered with Crossref DOIs. Though you can still create a relationship between the two DOIs if the content owner is aware through other means (e.g authors updating).
Given that we are seeing proliferation of preprints and different versions of papers available on the internet, being able to explictly identify and link versions of the same work is very important. While Unpaywall does do a good job to aggregate different versions together this is mostly based on heursitics such as similarity of title and authors.
Relationships can also be made between different research objects not just different versions of the same work.
The most obvious type of relationships is of course cites between two papers both of which have Crossref DOIs.
Technical note : You can create such citing relationship in two ways. The traditional way is via reference lists and the newer way is via a explict typed relationships.
Here's a less obvious example, as we saw above Crossref DOIs can be assigned for Peer reviews and we can link these peer review DOIs with the work they peer reviewed in VoR with the relationship isReviewOf
Lastly, one of the more important relationships that has gotten a lot of attention is the relationship between Datasets and published article aka Data Citation which can be connected with isSupplementTo , IsReferencedby etc
I won't go into detail here but essentially you can create such a relationship both from the Crossref side (usually Publisher) using two different methods (via references or relationships - the later allows nuanced relationships) or from the Datacite side (usually data repository). Assuming this is done, you can query such links via Crossref Events API or Datacite API. For details refer to this webinar or guidance from Crossref
So assuming all these relationships and links are created, what are the implications?
In 2016, Crossref talked about the idea of a Research Article Nexus defined as the sum of all these relationships between research objects and how they intend to expose all these on a open platform, free for all to use.
With all the relationships declared across all 80+ million Crossref metadata records, Crossref creates a global metadata graph across subject areas and disciplines that can be used by all.
It's important to note that when creating relationships (typed relationships) from Crossref DOis to other objects you are not linked to just other Crossref DOIs (or even Datacite DOIs). You can create relationships with DOIs from other RAs (such as Datacite) or any other Persistant identifer as Crossref Schema is accepting PIDs such as ORCID, ROR, ISBNs and more as controlled fields.
This has led to FREYA's concept of a PID graph, which is basically built on a similar idea but goes beyond Crossref DOIs by looking at a graph linked by PIDs including DOIs (various RAs), ORCIDs, ROR, ISBNS, ISSNs and more.
Datacite introduced this idea of the PID graph in 2019, which draws on the existing and growing PID infrastructure around research objects by creating a web/graph of PIDs by linking relationships between different PID types using various sources including the already mentioned Crossref events data.
https://www.crossref.org/documentation/content-registration/structural-metadata/relationships/
Using GraphQL (a query language designed for querying graphs) you can run queries like show me all datasets that are cited by publications that have authors affiliated with my institution. It may work like this
1. Look up authors by ORCID who are affiliated with your institution ROR.
2. Pull up all publications of such authors
3. Look at the data citations of such articles
4. Show datasets
Clearly there's a lot of potential if the PID graph takes off, though currently most PIDs are not that well linked yet.
Conclusion
This has been a very long blog post, and I just scratched the surface about the functionality and importance of DOIs and Crossref/Datacite. Still you can see how these infrastructure and their related services have implications on many topics of interest to librarians including
availability of open scholarly metadata with its various implications
Retractions and updates
Plagarism checks
OpenURL and link resolvers
Text-mining methods
Tracking relationships between preprints and Version of record
tracking funding information
data citations and altmetrics
and many more.
Each service or section deserves several blog posts devoted to it but this blog post was more to give you a taste of how far reaching the implications are to convince you the importance of why it is worth understanding DOIs.