non-OBO PURLs in Foundry ontologies

[wdduncan]

Making an issue for question asked on the google group.

I am working with a group that wants to use the Behavior Change and Intervention Ontology (BCIO) in conjunction with OBO ontologies (e.g., mental functioning ontology).

If they want to submit their ontology to the OBO Foundry, do we allow these ontologies to use non-OBO PURLs? For example, BCIO has the relation 'has study sample':
http://humanbehaviourchange.org/ontology/BCIOR_000001

Although BCIO is the specific use case, there are ontologies to consider (e.g., the Common Core Ontologies or Industrial Ontologies Foundry).

Possible principles for using non-OBO purls:

• The IRI must resolve in a web browser
• PURLs not to use their own domain name
• Some kind of good-faith but strong evidence that a bespoke URL base will continue to be maintained after funding ends

I'm on the fence about this. On one hand, I like it b/c it avoids OBO ontologies duplicating work done elsewhere. But, on the other hand, it may introduce a lot of crap into the OBO Foundry space.

Perhaps we classify ontologies using non-OBO PURLs as so-called "application ontologies". I'm also in favor of permitting application ontologies to inject axioms (see discussion here).

cc @cmungall @matentzn @addiehl

[cmungall]

I don't think use of non-OBO PURLs is a deal-breaker, but with my non-OBO hat I would strongly recommend that groups do not use their own domain name in PURLs or bespoke URLs. Use an existing system like purl.org, w3id.org, or purl.obolibrary.org. For OBO, if we do allow non-OBO namespaces (see discussion on FMA), I recommend some kind of good-faith but strong evidence that a bespoke URL base will continue to be maintained after funding ends (and that good practice will be followed while you have funding). I think this is just good broad semweb practice, nothing to do with OBO per se. I would be curious to see what the reasons are for needing a bespoke URL

On Sat, Dec 21, 2024 at 11:46 AM Bill Duncan ***@***.***> wrote: Making an issue for question asked on the google group

<https://groups.google.com/g/obo-operations-committee/c/iCqfKvO6TeI/m/J3n3CLL5AAAJ?utm_medium=email&utm_source=footer&pli=1> . I am working with a group that wants to use the Behavior Change and Intervention Ontology (BCIO) <https://www.bciontology.org/> in conjunction with OBO ontologies (e.g., mental functioning ontology). If they want to submit their ontology to the OBO Foundry, do we allow these ontologies to use non-OBO PURLs? For example, BCIO has the relation 'has study sample': http://humanbehaviourchange.org/ontology/BCIOR_000001 Although BCIO is the specific use case, there are ontologies to consider (e.g., the Common Core Ontologies or Industrial Ontologies Foundry). Possible principles for using non-OBO purls: - The IRI must resolve in a web browser - PURLs not to use their own domain name - Some kind of good-faith but strong evidence that a bespoke URL base will continue to be maintained after funding ends I'm on the fence about this. On one hand, I like it b/c it avoids OBO ontologies duplicating work done elsewhere. But, on the other hand, it may introduce a lot of crap into the OBO Foundry space. Perhaps we classify ontologies using non-OBO PURLs as so-called "application ontologies". I'm also in favor of permitting application ontologies to inject axioms (see discussion here <#1443 (comment)> ). cc @cmungall <https://github.com/cmungall> @matentzn <https://github.com/matentzn> @addiehl <https://github.com/addiehl> —

…

Reply to this email directly, view it on GitHub <#2667>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAAMMONBWR74TIBYYLOHE732GXASBAVCNFSM6AAAAABUA4XAG6VHI2DSMVQWIX3LMV43ASLTON2WKOZSG42TIMZTGQYTENI> . You are receiving this because you were mentioned.Message ID: ***@***.***>

[wdduncan]

I would strongly recommend that groups do not use their own domain name
in PURLs
Interestingly, the BCIO is part of a larger effort: The Behavioural and Social Sciences Ontology (BSSO) Foundry

The BSSOF includes many namespaces.

I recommend some kind of good-faith but strong evidence that a bespoke URL
base will continue to be maintained after funding ends

I agree. I suppose we need to assess what counts as "good-faith but strong evidence".

[ddooley]

I'll say that a mini-project in an upcoming FoodOn hackathon is to create an RDF graph of geographical names and some geographical relations extracted from Wikidata/Wikipedia for use within FoodOn and GenEpiO, taking over from past use of GAZ. So this is a case where non-OBO purls will be offered to all on a script-driven github site I think. Hope that goes down well! The Wikidata "slim" will be offered to all. If this seems problematic let me know!

[cmungall]

That sounds eminently sensible @ddooley. I don't know if there is a way for OBO to be meaningfully involved, this is more like what many of us do for conversion of biological knowledge bases

See:

• How should we standardise the use of database identifiers in OWL Axioms? #1981
• https://github.com/biopragmatics/obo-db-ingest
• https://docs.google.com/presentation/d/1aySEHTgkags7UPJYHyvQ9frYvAIqr1G5A3u7dGF26Y4/edit#slide=id.p

[matentzn]

Here is an importat warning story why never to use simple URLs for domain without a long term maintenance plan for IRIs:

#2425

[wdduncan]

@matentzn Thanks for the warning. However, I am not sure how to heed the warning. What counts as a "long term maintenance plan"? I would imagine there are different thoughts on what counts as such.

I ask again for folks to consider allowing non-OBO PURLs in what I've calling "application ontologies" (or some other label). These are in my mind ontologies that have less restrictions, such as allowing axiom injections and non-OBO PURLs.

[matentzn]

However, I am not sure how to heed the warning. What counts as a "long term maintenance plan"? I would imagine there are different thoughts on what counts as such.

There is a simple criterion that is not purpose: someone intends for a domain to serve as a PURL server, and presents a sustainability plan. w3id.org and obolibrary.org have done so. The sustainability plans are explicit and exist on the web.

I ask again for folks to consider allowing non-OBO PURLs in what I've calling "application ontologies" (or some other label). These are in my mind ontologies that have less restrictions, such as allowing axiom injections and non-OBO PURLs.

If the classes are not intended for re-use, anything is fine of course! If they are, then in my view, no, for the reasons outlined above.

[wdduncan]

There is a simple criterion that is not purpose: someone intends for a domain to serve as a PURL server, and presents a sustainability plan. w3id.org and obolibrary.org have done so. The sustainability plans are explicit and exist on the web.

I think that may be too high of a bar. If the ontology is published on a public ontology browsing service (e.g., Bioportal), I think that is sufficient. The IRIs are able to be looked up, and since they are available on a public server that should cover the reuse question.

Others should weigh in, though.

[cmungall]

I think that may be too high of a bar.

How so? Surely it's less of a high bar that running your own webserver and ensuring that you have a plan to renew domains continually etc

If the ontology is published on a public ontology browsing service (e.g., Bioportal), I think that is sufficient.

Great example! I think that actually illustrates the point @matentzn and trying to make. There is a bioportal/ontoportal sustainability plan which I can share, I would personally consider bioportal PURLs as good choices from a PURL sustainability POV.

Note that I am still speaking from a common sense general OWL / semantic web perspective and not on OBO policies.

[wdduncan]

I think that actually illustrates the point @matentzn and trying to make. There is a bioportal/ontoportal sustainability plan which I can share, I would personally consider bioportal PURLs as good choices from a PURL sustainability POV.

Sorry. I'm not following you about this point. Bioportal doesn't have a PURL service. IRIs that being with http://purl.bioontology.org/ontology/ (e.g., http://purl.bioontology.org/ontology/RCTV2/2JQ..00) do not resolve.

I recently enquired about whether they (Bioportal) could implement a PURL service, and I was told that were not planning to do this. I can forward the email if you wish. Perhaps you mean something different?

How so? Surely it's less of a high bar that running your own webserver and ensuring that you have a plan to renew domains continually etc

If I understand you, this would require that the ontologies use w3id.org, purl.obolibrary.org, or some other redirect service in their IRI. This would exclude many of the ontologies on Bioportal. I would also exclude (at the moment) terms from the Common Core Ontologies. Surely, there are many ontologies with useful terms whose IRIs don't resolve. I think this too strict of a standard.

[nataled]

It looks like that particular PURL example is not well formed. Here's another that does resolve: http://purl.bioontology.org/ontology/RCD/X78tJ. That being said, it looks like these are not customizable (that is, can be configured to redirect) in the same way Foundry PURLs are.

I agree with @wdduncan regarding IRIs that don't resolve. My understanding from past discussions (way way past!) is that IRIs are, first and foremost, identifiers. We have no language that requires IRIs for individual terms to resolve, even though the IRI for the ontology as a whole MUST resolve. Instead we say this on our ID policy page:

The URIs should resolve to useful information about a term.

The word used is 'should', not 'must' (This page was created before we routinely made such words all uppercase).

[wdduncan]

Thanks @nataled ! I was not aware of well-formed issue with bioontology.org PURLS.
Not wanting to derail the conversation. But, is there a page about what format the bioontoogy PURLS need to have?

[jamesaoverton]

The URIs should resolve to useful information about a term.

Even though the current language is "should", it's a MUST for me. "Findable" is the "F" in "FAIR", and it's table stakes for ontology use and reuse.

[cmungall]

Apologies, I assumed the bioportal PURLs resolved. Let's take this one up on the bioportal tracker as it's a diversion from the main topic at hand here.

I agree with @jamesaoverton re MUST. And to restate the position @matentzn and I are articulating, for the "P" is PURL:

There is a simple criterion that is not purpose: someone intends for a domain to serve as a PURL server, and presents a sustainability plan.

[wdduncan]

Apologies, I assumed the bioportal PURLs resolved. Let's take this one up on the bioportal tracker as it's a diversion from the main topic at hand here.

Sure. I was just being curious if there was documentation.

So, we are at a crossroads. I think the terms are findable enough if you can look them up on a service like Bioportal. But, I may be in the minority about this opinion.