Review 4
-
Reasons to reject:
There are a number of serious inconsistencies within the manuscript that raise significant questions about the manuscripts readiness for publication:
-
Conflating persistence of URIs with link rot
"The author states that “The prevalence of phenomena such as link rot implies that the persistence of URIs cannot be trusted.” The persistence of URIs is independent of link rot – just as the existence of an ISBN or DOI doesn’t guarantee that the book/paper is available. For digital identifiers, this is resolved through the use of tombstone pages."
-
Conflating PIDs with Open Access
"The authors state that “another reason resolvability may not be sufficient, even if the metadata is somehow in place, is that the file on the destination page resolved to is behind a paywall. In a case from December 2016, public domain content more than 110 years old was hidden behind a DOI-resolver charging 50$ for release of the content.” PIDs do not guarantee open access and shouldn't. In addition to potential costs of material (as noted above) there are many cases where content identified by a PID should not be available without access restrictions (e.g. Protected Health Information). For the examples given in the manuscript, the PIDs described in the manuscript do provide metadata for the resources identified. The authors continue to discuss this further “However, things have changed since then. When tried again now (Nov. 2018), the replacement unpaywall.org [22] and oadoi-API for 10.1080/00222930908692639 is actually not working anymore for this DOI; the response we get is: best_oa_location: null. But the resource sought, free from paywall, although no longer detected by unpaywall.org, can still be found at biodiversitylibrary.org, at several different URLs.” PIDs are needed to identify open content and content that has various restrictions (and not only restrictions due to cost). PIDs should not be a guarantee of open access to underlying content – however all PIDs should have appropriate metadata (see 3 below)."
-
Don’t need to jump to landing page for metadata:
"The authors state that “Again, going back to the question of resolvability, the relationship between identifiers such as DOIs and URIs/IRIs is not always straightforward, and sometimes involves a chain of redirects ('303s'), before reaching eventually a destination holding also the appropriate metadata.” However, that is not the case as the metadata for any DOI is easily available: https://search.crossref.org/?q=10.1080%2F00222930908692639 VII.—Descriptions of new genera and species of New-Zealand Coleoptera Journal Article published Jul 1909 in Annals and Magazine of Natural History volume 4 issue 19 on pages 51 to 71 Authors: Major T. Broun And metadata for un-resolvable DOI exists: Showing DOI matching 10.1002/(sici)1520-6297(199601/02)12:1<67::aid-agr6>3.3.co;2-# Acreage response under policy incompatibilities: The US durum wheat situation Journal Article published Jan 1996 in Agribusiness volume 12 issue 1 on pages 67 to 77 Authors: Roberto J. Garcia, James E. Quinton"
-
Potential issues with identifier explosion
"The authors contend that “Providing multiple access to, or identification of resources through PIDs, that are capable of serving as trustworthy, competent, valid independent witnesses from different moments in time, at different sites, in different places is a good idea. Thus, we accept “that an object may have multiple PIDs”. Ideally these multiple PIDs should get to "know about" each other as a way towards interoperability.” This explosion in identifiers potentially poses significant problems. A primary issue would be the systems required to keep the synonyms in check – the overall ecosystem would need another PID-like service that maintains this information. In addition, there already exist independent and trustworthy resources that resolve PIDs. For example, for DOIs the resources dx.DOI, n2t.net and identifiers.org (and others) all resolve metadata for DOIs."
-
Incorrect description of DOI requirements
"The authors state that “according to the same partial restriction, this entirely fake DOI is equally valid: 10.99999999/xxxxxxxx/x(y)x\:-{=?%%@@@@@“ However, this is incorrect. The initial component of the DOI (10.99999999) is organization that controls the ID space and this is not a valid DOI prefix."
-
Misstating the lack of usefulness of certain IDs
"The authors state that “UUIDs generated in this way by the gna name resolver, e.g. "707f84e1-e5b8-5063-8256-369ba9d72e13" for Antiaris toxicaria are next to useless as instruments of Findability, often yielding 0 hits by simple googling, all the while a search on the scientific name alone will give plenty of precision hits for the sought after organism, providing rich metadata for the 'thing' itself.” First this is not the issue with the PID itself - rather its utilization. There are many PIDs that have very good findability via Google as their use entails incorporation in the text of documents (e.g. database accession numbers from PDB or GEO, the use of RRIDs in methods sections). In addition, googling for names does not always produce proper results – a primary reason for the increased interest in the use of ORCIDs to enable author disambiguation."
-
Model suggested is not differentiated from current models
"The authors propose a model for PIDS: “Model: [namespacePrefix].[objectType].[objectId: 10 positions].[issuedDate: YYYY-MM-DD].[registrant: org.id/ORCID] Example (expression of this paper): fabio.PositionPaper.jPsaveXD17.2018-11-12.0000-0001-5699-994X” However, this model is very similar to current handles. It seems that the two pieces that determine the “prefix” that is who created it and for what purpose are split. In the proposed model. However, this would be similar to allowing anyone to create a handle with derived prefixes which is allowed by the handle infrastructure: “The GHR is a distributed registry whose operation is managed collaboratively by the DONA Foundation and multiple organizations that are credentialed and authorized by DONA. Each of those credentialed organizations are referred to as MPAs (or Multi-Primary Administrators). A credential is a number, made known to the GHR by DONA and allotted to a given MPA. MPAs, in turn, are authorized to allot derived prefixes from their credential to themselves and to third parties. The MPA and these third parties can provide identifier and resolution services (aka local handle services) for handles under the derived prefixes allotted to them. The GHR specifically communicates to the client, upon request, the location and certain relevant security information of the local handle services that can process the handle that the client is interested in resolving.” The authors also discuss that the “resulting PIDs should be minted within the corresponding namespaces, who would also be the 'custodians' and resolving authorities of their PIDs, responsible for their uniqueness within their namespace. Another task would be to monitor and assign sameAs-properties to PIDs that refer to the same 'thing' in other namespaces.” Would this then result in a collection of un-coordinated custodians? This is an issue with database accession numbers and has required construction of additional infrastructure to support the resolution of these PIDs. Coordination and management of the sameAs-properties cold become problematic as well and isn’t discussed."
-
-
Response:
-
I am not sure I understand this objection. Link rot does sometimes prevent using a URI for direct identification of an object, and that is precisely what identifiers are for. There are not always "tombstone pages" available, and even when there are, as in the cases referred to in objection 3, for 10.1002/(SICI)1097-4571(199510)46:9<646::AID-ASI2>3.0.CO;2-1 and 10.1007/s11192-007-1682-3, these are not so "easily available" as being referred to directly from the destination page of the DOI-URI, where you are only met by the message "DOI Not Found" or "Page not found". This means at least that the PID-URI in these cases cannot be considered to be properly machine-actionable, a pre-requisite for FAIRness; unless you implement an automatic redirect in your workflow for these cases, you have to make a new manual search using the Crossref search-API https://search.crossref.org/?q=, something that may not be obvious to everyone using PIDs (including this author, who is thankful to the Reviewer for this piece of information). Of course, you could say that the PID-URIs are still persistent, while the links are not, but that is a mere play with words. The purpose of a URI = Uniform Resource Identifier, is to identify an object, an entity, and if the link does not work any longer, it simply fails duty. It is for this very reason that the paper argues for PIDs that are well distributed over the Internet, 'googlable' and not dependent on one single link-URI, one single custodian or "parent" to fulfil their mission. It is in this respect that ISBNs serve as an example, not because you will always find a copy of the actual book that they identify - it may be on loan at your local library or out of print, not in store, but because you will most often be able to identify the object that they are supposed to identify, through many online library catalogs, bookstores, reference lists, net resources etc.
-
I agree with the view that "PIDs do not guarantee open access and shouldn't." However, we are dicussing here the A in FAIR, Accessibility, and I do not think it should be the purpose of PIDs to make content that belongs in the public domain in fact less accessible, as in the case referred to here. For the rest, see my response to review 2 above.
-
First, it should be noted that the introductory quoted sentence in the remark beginning "Again, going back to the question of resolvability ..." has two references, , , representing Crossref and DOI, so this statement is not something this author claims originality for. For the rest, see my response to Review 4, point 1 above.
-
The final sentences under this point puts in doubt if the reviewer has actually read carefully the paper under review, in which, under section 4, is found the following information:
Note that we are not talking here about simply having more than one proxy server acting as resolvers of the same PIDs. We already have that; provided the lookup-table is managed properly, the three different DOI-URIs from three different proxy-servers all resolve to the same landing-page location: https://doi.org/10.1007/978-3-319-53637-8_11, https://hdl.handle.net/10.1007/978-3-319-53637-8_11 and https://identifiers.org/doi:10.1007/978-3-319-53637-8_11. ARKs (Archival Resource Keys) are resolved by identifiers.org and n2t.net, as well as by their "mother institutions", e.g. n2t.net/ark:/67531/metapth346793/, identifiers.org/ark:/67531/metapth346793/ and digital.library.unt.edu/ark:/67531/metapth346793/ resolve the same content.
As for the propositions quoted in the beginning of the reviewer's point 4, they all lean on as their main reference , a presentation by Jonathan Clark, representing the International DOI Foundation, at the first PIDapalooza in Reykjavik 2016. So, again, this does not represent a major invention by this author. And if you read further from there, you will also find a reference, again, to the NUNA, the Non-unique Naming Assumption. Maybe this is is not enough to dissipate the fear of an "identifier explosion", that I believe already happened, but I do agree that we need more agents, not only one central PID-like service to "keep synonyms in check", and we need it already today. The paper refers to the SPARQL endpoint of identifiers.org as one such important agent, but this should be a common responsibility of all those namespace custodians who want to mint PIDs of their own, thus to contribute to the FAIR effort of making PIDs, as an essential part of metadata, Re-usable also in the sense that they are "sufficiently well-described and rich that it can be automatically (or with minimal human effort) linked or integrated, like-with-like, with other data sources" (FAIR Guiding Principles 4.2). Perhaps, this responsibility should also make them avoid minting new PIDs for the same 'things' in vain, when there are already sufficiently well-known, well described, 'validatable' PIDs in widespread use for the same objects.
-
Again, the impression is received that the reviewer has not read the paper under review sufficiently carefully to see that "the same partial restriction" here naturally refers back to the preceding regex above (to make this evident it is spelled out in the revised version of the paper submitted with this response to reviews). And the restriction by this regex fully matches with this fake DOI, no matter what the prefix rules for DOIs say, as shown below.
Furthermore, unfortunately, the same goes for the even less restrictive rules for DOI validation used by DataCite "10\..+/.+" , or the pattern registered for DOIs at identifiers.org as "^(doi\:)?\d{2}\.\d{4}.*$" MIR:00000019, both of which also allow for the fake DOI above as valid, when tested in regex101.com. It is of course possible to construct a more restrictive regex that rules out false DOI prefixes, and thus would have this fake DOI fail to validate, but that obviously remains to be implemented yet in validation schemas used by important agents such as DataCite and identifiers.org. And, as is further argued in the paper, given the quite lax initial rules for DOI structure, once you try to be more restrictive, you obviously will not catch all the now prevalent and permitted DOIs by one singular regular expression. , So, as is stated in the paper, "the more restrictive the validation rule or regular expression it is based on, the more actually existing DOIs it will leave out."
-
One reason for the uselessness of the UUIDs issued by the gna name resolver as an instrument of Findability for the organism, the 'thing itself' is clearly stated in the paper in the sentences just preceding the one quoted by the reviewer: "Note, however, that it is actually the specific name string that is identified here, not the object, the organism, the 'thing itself'. Thus, the resulting UUID is completely dependent upon the particular name string (with its encoding), it cannot be used as a bridge between different name forms for the same organism, telling us that they are naming the same object. This is due to the fact that it is
generated by hashing a namespace identifier and name
. " This is what thein this way
in the sentence quoted (in part) by the reviewer refers to. But I agree that the uselessness of the UUIDs in this respect is at least partly due also to it's not being further utilized for citations or disseminated to other databases, while suspecting this fact is closely related to the preceding way in which they are created as hashes of name strings, rather than as identifiers of objects, concepts, 'things'. The point I am trying to make, with reference again to is that PIDs must be used to be persistent, and that the chances of becoming used and utilized for identification are better if the PID by it's own structure, revealing some semantic content, somehow reflects what kind of 'thing' it is supposed to identify. In this sense, I believe a quite long, totally opaque string, such as the UUID-5, that on the face of it appears to be more or less random (although in reality there is a strict hashing algorithm behind it), might possibly be less successful in this respect. I know there are other databases, with identifiers, often shorter, such as accession numbers, that are more successful in this respect, e.g. UniProt, whose PIDs are also resolved by identifiers.org, but then only if you use the right namespace prefix, e.g. http://identifiers.org/UNIPROT:P68512, which already tells you something about what kind of 'thing' you are supposed to identify, i.e. having some semantic content. One might add that databases such as UniProt are known to have made an extra effort to be findable by 'googling' through the use of schema.org markup. This is something that should be encouraged in all PID-custodians or producers, as a way of disseminating and enhancing the use of these PIDs! Naturally, 'googling' is not the only solution to enhanced findability, and certainly googling for names does not always produce good results. In particular, in section 2 of the paper it is argued that "While scientific names are often useful for describing objects, they have other drawbacks compared to PIDs, some of which were identified by . For example, homonymy and disambiguation should generally be a lesser problem for globally unique identifiers." So, this remark about ORCIDs seems a little bit misplaced or at least redundant here. Please note also, that it is for a reason, i.e. precisely the awareness of the sometimes dubious results of 'googling', that the estimated precision rates were included in some of the examples given in the paper. (Recall rate, unfortunately, is harder, perhaps even impossible, to estimate in these cases.) -
This objection is also a bit difficult to understand. First, the paper makes no specific claim of originality in this respect. That is why in the beginning of section 7 the word model is emphasized and there are quotation marks surrounding "new" (now in the revised version also in the section heading, to make it clearer):
So, here we finally suggest a model for a "new" PID, with a limited character set, at least for the object id part, defined by namespace specifications and schemas.
The "mission" of the paper is simply to explore the concept of a PID, looking at real examples, their positive and negative features, and find out what additional requirements there might be to make them fully findable, accessible, interoperable and re-usable - FAIR. The "novelty" of the paper, if any, would then rather be the "widening" of the FAIR principles to include also Findability as rate of distribution or dissemination (measured by means of 'googling') and Interoperability or Re-usability to include also 'validatability'. As for the custodianship and minting of PIDs, the model proposed is that this should be the responsibility of the namespaces to which they belong. I do not see why these, already assuming the administration of specifications, validation schemas, vocabularies or ontologies, should be any less qualified for this task, than the MPAs in the Handle system. The minting algorithm, the patterns for PID-recognition, restriction in character set, string-length (with possible checksum) of objectId module should all be part of the validation schema. These namespaces should then be able to register their schemes with n2t.net or identifiers.org, as already happens. So, why should the danger of "a collection of un-coordinated custodians" be any more real than today? Besides, PIDs are meant to work on a network, essentially by means of links between nodes in the network. Have we not come further beyond the idea of a central authority ruling everything? Possibly many different actors and namespace custodians could contribute to creating "sameAs" links between PIDs identifying the same 'things'. And there might be several services such as the SPARQL endpoint of identifiers.org for registering such links.
-
