> By the way I am noting that LoC's plan for migration to 
> HTTPs seem to exclude the ontologies/schemas part of
> LoC's offer. Does this reflect a policy?

-- Oddly enough, it has to do, in part, with the very reservation you have expressed.  Fear of breaking things.  The other aspect is that, again generally speaking, there has been a conscious effort not to alter the URIs of classes and properties (as published by ontologies/vocabularies).  You note DCMI, but others too.  Since it seemed that the community was avoiding change when it came to classes and properties, we were going in that direction (see earlier statement about trying to align what we do here with what others are doing).

Yours,
Kevin


-----Original Message-----
From: LC Linked Data Service Discussion List <[log in to unmask]> On Behalf Of Antoine Isaac
Sent: Wednesday, June 17, 2020 10:23 AM
To: [log in to unmask]
Subject: Re: [ID.LOC.GOV] ID switching to HTTPS

Thanks for the answer, Kevin.
And I'm glad Osma has chimed in.

I can understand the concerns expressed at SWIB. But still I have doubts about some points, especially that the Semantic Web is still young and can afford changing its URIs. I wonder what would happen if DCMI would decide to change the identifiers for DC terms from HTTP to HTTPs...
  By the way I am noting that LoC's plan for migration to HTTPs seem to exclude the ontologies/schemas part of LoC's offer. Does this reflect a policy?

Cheers,

Antoine

On 17/06/2020 15:34, Osma Suominen wrote:
> Hi Kevin,
> 
> I have the same concern as Antoine.
> 
> In our YSO (General Finnish Ontology) data, we have more than 10k statements similar to this:
> 
> yso:p1191 skos:closeMatch <http://id.loc.gov/authorities/subjects/sh85145789> .
> 
> which is machine-speak for "the YSO concept of 'waves' is pretty much the same as the LCSH concept 'Waves'".
> 
> Then we use our own Skosmos software to display this information in Finto.fi. Skosmos performs a lookup to the LCSH URI above, parses the RDF it gets back, then tries to find a triple like this in the data:
> 
> <http://id.loc.gov/authorities/subjects/sh85145789> skos:prefLabel ?label .
> 
> And this matches the label "Waves"@en, and this gets displayed in the UI, and all is well.
> 
> 
> But, if you switch to HTTPS, as it currently stands, Skosmos doesn't understand the RDF response you seem to be aiming for:
> 
> <http://id.loc.gov/authorities/subjects/sh85145789> owl:sameAs <https://id.loc.gov/authorities/subjects/sh85145789> .
> 
> <https://id.loc.gov/authorities/subjects/sh85145789> skos:prefLabel ?label "Waves"@en .
> 
> 
> Now this can be fixed in many ways (including changing Skosmos to 
> follow owl:sameAs statements), and most likely if the canonical LCSH 
> URIs switched to HTTPS, we would make the same change for all the YSO 
> statements, and everything would sync up nicely again. But requiring 
> everyone to change their URI references like this is a bit problematic 
> IMHO. For us it's probably quite easy, since the LCSH URIs are used in 
> only one data set (YSO) and only for mappings, not in e.g. 
> bibliographic databases with millions of records. (Though we do have 
> some references to other LC vocabularies as well in a few places)
> 
> One somewhat less disruptive change would be to keep the HTTP URI as the canonical ones (with the label statements etc.), but still use owl:sameAs to point to the HTTPS version, like this:
> 
> <https://id.loc.gov/authorities/subjects/sh85145789> owl:sameAs <http://id.loc.gov/authorities/subjects/sh85145789> .
> 
> <http://id.loc.gov/authorities/subjects/sh85145789> skos:prefLabel ?label "Waves"@en .
> 
> This would not retract any currently available triples from the data set, just add a bunch more.
> 
> -Osma
> 
> PS. Actually I simplified the picture a little. For many years we did traditional Linked Data URI lookups of the kind I explained above to LCSH, but those were quite slow and unreliable, so in Skosmos 2.5 (released April 2020) we switched to the ad-hoc X-PrefLabel HTTP header mechanism that id.loc.gov provides. So the above is no longer accurate in the Finto/Skosmos/LCSH case, but other Linked Data resolvers will probably do things in a similar way.
> 
> Ford, Kevin kirjoitti 17.6.2020 klo 16.07:
>> Thanks, Antoine, for the info.
>>
>> Nothing has changed, yet, so your input is helpful.  We recognize the issue you raise - we have it too since we use our own URIs everywhere.
>>
>> To be honest, we're using this informative page as a starting point:
>>
>> https://wiki.dnb.de/display/DINIAGKIM/HTTP+vs.+HTTPS+in+resource+iden
>> tification
>>
>> I was not at SWIB so I did not participate in the discussion.  As 
>> such, I don't have firsthand experience with the discussion and 
>> therefore anything that might have been accidentally omitted from the 
>> notes.  My general takeaway is that 1) there is still a number of 
>> questions surrounding this but 2) there seemed to be a general 
>> movement toward changing the URIs to HTTPS and employing an 
>> owl:sameAs relationship.  That seems to have been the case with 
>> Sweden, certainly.  And BNF and DNB use HTTPS URIs though I can't 
>> seem to find - or am unclear about - the owl:sameAs assertions for 
>> those two libraries.  (It could be I was looking in the wrong place.)
>>
>> We're mostly interested in going in a direction that will cause least friction and align with the general trajectory of others facing this similar issue.  This is why your example is helpful, since it departs from what we largely see the community doing from the notes of that SWIB meeting.
>>
>> Do others have thoughts?
>>
>> Yours,
>> Kevin
>>
>>
>>
>> -----Original Message-----
>> From: LC Linked Data Service Discussion List <[log in to unmask]> On 
>> Behalf Of Antoine Isaac
>> Sent: Wednesday, June 17, 2020 4:55 AM
>> To: [log in to unmask]
>> Subject: Re: [ID.LOC.GOV] ID switching to HTTPS
>>
>> Hello Kevin,
>>
>> Thanks for the heads-up!
>>
>> I think I will ask for a double-check: the vocabulary URIs like LCSH will now de-refer to data where the statements will use HTTPs URIs?
>>
>> Using a toy example...
>>
>> If Europeana receives statements like ex:book dcterms:subject 
>> <http://id.loc.gov/x>
>>
>> Then when getting the data for <http://id.loc.gov/x> we will get these statements:
>> <https://id.loc.gov/x> skos:prefLabel "label" .
>> <https://id.loc.gov/x> owl:sameAs <http://id.loc.gov/x> .
>>
>> Is that correct?
>>
>> If yes, it means that data users will have to rely on owl:sameAs machinery or re-write all URIs, otherwise their data will be 'broken' (i.e. they wouldn't directly get statements about the URIs that they look up).
>>
>> For the record at rightsstatements.org we have considered the issue 
>> and opted to still use HTTP URIs as "canonical" identifiers in the 
>> data we serve (though there's HTTPs in the service that provides the 
>> data and web pages). And we hope that other actors like Creative 
>> Commons will *not* make the move of using HTTPs URIs as "canonical" 
>> identifiers, because that would force us to make too many changes 
>> (also in non-RDF aware part of the software, and documentation!)
>>
>> All the best,
>>
>> Antoine
>> ---
>> Antoine Isaac
>> R&D Manager, Europeana
>>
>> On 16/06/2020 23:46, Ford, Kevin wrote:
>>> Most faithful users of ID.LOC.GOV:
>>>
>>> In the next 3-4 weeks, ID will be switching to HTTPS.  Until now, ID has operated via HTTP, though for several years now HTTPS has been an option for anyone who wanted to use HTTPS instead.
>>>
>>> A 301 (Permanent Redirect) will be implemented to direct HTTP requests to their HTTPS equivalents. There is no end date at this time for the 301 redirect, but we encourage everyone to start using the HTTPS URIs once the switchover happens. It'll decrease latency (fewer HTTP calls/redirects) and, of course, ensure your communication with the service is encrypted from the outset.
>>>
>>> Except for the classes and properties of ontologies/vocabularies hosted at ID - Bibframe, MADS/RDF, and PREMIS - the HTTPS URI will become /the/ URI for resources.  We will add owl:sameAs properties to those resources to capture the fact that the HTTPS URI is the same as the HTTP URI.  To be clear: the URIs for classes and properties of ontologies/vocabularies hosted at ID will still be referenced as HTTP URIs even if the client is redirected to an HTTPS address.
>>>
>>> The URIs in the bulk downloads are one of the trickier aspects of this change.  Our thinking presently is to re-issue all the bulk downloads as soon as we can wherein the HTTPS URI is /the/ identifier for resources.  What we're not clear about is how disruptive this will be for those of you who make regular use of the bulk downloads.  We'd welcome your feedback either way as this will help us determine if re-issuing the bulk downloads with the new URIs is acceptable or whether we need to explore additional options.
>>>
>>> Do let us know.
>>>
>>> All the best,
>>>
>>> Kevin
>>>
>>> --
>>>
>>> Kevin Ford
>>>
>>> Library of Congress
>>>
>>> Washington, DC
>>>
>