W3C home > Mailing lists > Public > www-tag@w3.org > June 2021

Re: Principles of Identity in Web Architecture

From: Henry Story <henry.story@bblfish.net>
Date: Wed, 16 Jun 2021 15:34:41 +0200
Message-Id: <B436DC19-D603-4DA1-A03D-9936A71E808E@bblfish.net>
Cc: "www-tag@w3.org" <www-tag@w3.org>, Peter F Patel-Schneider <pfpschneider@gmail.com>
To: Philip Sheldrake <philip@eulerpartners.com>


> On 16. Jun 2021, at 10:14, Philip Sheldrake <philip@eulerpartners.com> wrote:
> 
> 
> It seems, following Henry’s introduction of HoTT here, that we need to add mathematicians to that list of disciplines to have around the table!

There are a few lurking. I spent the last 5 years studying Category Theory and was thinking
of getting a  Web-Cats Community Group going. It has a web page:

   https://web-cats.gitlab.io

There is a fun video there about Cat Herding that should feel very relevant to members of the TAG.
(Still I am not certified, and I am only at the beginning of exploring these topics, and have run
out of money doing that)

> 
> As Melvin picked up from generative-identity.org, we (AKASHA Foundation) think it useful in this context to consider conceptualizations of identity as noun-like or verb-like (per Buckminster Fuller’s “I Seem To Be A Verb.”)
> 
> The primary characteristic of verb-like conceptualizations is their accommodating and enabling change; hence the direct consequences for psychological, sociological, and ecological health. Who I am encompasses a constant flux of informational diffusion and intermixing, interfacial constructions and experiences, continuously revised narratives, arrangings and organizings. At least that’s my way of putting it. Professor Kathleen Wallace is far more eloquent in this essay — You are a network. https://aeon.co/essays/the-self-is-not-singular-but-a-fluid-network-of-identities

That is a very nicely written essay on "personal" identities as revealed in a network of social relations to others.

I don't see a problem with the essentialist view, the processual view and the network view working
together.

The modern essentialist view was famously defended by Kripke in his lectures/book "Naming and Necessity”.
That is the biological processual view: you are the process that starts with the fertilization of a specific egg.
Processes in category theory are coalgebraic, and so are modal logics.

"Modal Logics are Coalgebraic"
https://www.dfki.de/web/forschung/projekte-publikationen/publikationen-uebersicht/publikation/5351/

This explains precisely why we can speak of essences here. The process that a person is can develop and interact in many different ways with the world giving us the ability to talk about different possible lives someone may have had. Different choices of action or differences in the world lead to different relations that a person could have had with others of different skills she may have gained or may lose due to war or scientific crazes like electroshock therapy.

Note that Coalgebras are the mathematical dual of algebras. And Algebras, in the most abstract sense, are the study of language like things.
https://ncatlab.org/nlab/show/algebra

> I know I’m repeating myself here when I note that verb-like identity conceptualizations are edge-centric rather than node-centric. It was me Henry who noted earlier here that the designers of the Internet and the WWW did not include human identity in their schema. And URIs — most evidently as URLs — are very much about the node and not the edge by my understanding. They address things, resources. Others here have far deeper understanding about such things (no pun intended), so do please correct me if I have erred.

This general idea you are putting forward seems very similar to the famous Yoneda Lemma from Category Theory which states that "mathematical objects are completely determined by their relationships to other objects.”
https://www.math3ma.com/blog/the-yoneda-perspective

This has been claimed to be "arguably the most important result in category theory”, so it is worth thinking about. Of course objects in the world are not only mathematical objects, and we need to take into account epistemological constraints: namely that we only have very partial views on reality.

But say that this captures your intuitions. It is not clear if this is problematic for the web or the semantic web.

For one: a very good description of the Semantic Web has been given in terms of bicategories of relations.
https://arxiv.org/abs/1706.00526

Bicategories are categories in which there are not just relations between objects, but also relations between relations.
In this case if there is one such, there there is precisely one such relation: the subPropertyOf relation.
(Beyond this are categories with relations between relations between relations up to infinity.)

I believe the Yoneda lemma applies there too.

Now for URIs. I discovered recently an important relation of the open world assumption to
natural living systems and (I believe) to URIs, as well as to Linked Data.

To start you should listen to this excellent non mathematical recent introductory talk

"Symmetric Monoidal Categories: A Rosetta Stone" by John Baez
https://twitter.com/bblfish/status/1397962748429062145

He argues that these type of categories are what is needed to describe open systems that can
be work together, such as electrical circuits, biological systems, etc…

For a 2 page definition of open systems you can read the first 2 pages of the
Preface of the Phd thesis of his student Brendan Fong which I uploaded on Twitter
with links here
https://twitter.com/bblfish/status/1398230682237911041

Now if you put that together with the paper by Peter F. Patel-Schneider
”Integrity Constraints for Linked Data” from 2011 which gives a model of Linked Data usage of
RDF should be modeled as a symmetric monoidial category you can start seeing exactly
what the roles of URIs are: they are what allows one to connect data produced by different
agents together in an open system.

I was reading the proof in Agda of the paper https://github.com/bblfish/agda-web-semantic
6 months ago, but then had to go off an work on an EU project for access control on the web.

https://nlnet.nl/project/SolidControl/

Henry

> 
> Verb-like conceptualizations don’t appear to attract much attention in the design of digital systems, perhaps because, while being dominant in our everyday lives, we don’t ‘see’ them in operation and / or no-one has yet perceived ROI in designing for them.
> 
> Noun-like identity conceptualizations on the other hand lack adaptability by design. The comparatively recent bureaucratisation of identity (legal identity) is noun-like.
> 
> In my first public critique of SSI a couple of years ago, I write:
> 
> “Surely one of the biggest question marks must be the ease with which noun-like and verb-like conceptualisations might co-exist. Perhaps there's a future in which only the latter are necessary, but a co-existence will be required for some good time to get us that far.
> 
> “It could be that the noun-like require entirely different protocols to the verb-like, and rigorous effort is made to design appropriate and enduring sociotechnological constraints to prevent the otherwise inevitable noun-like creep.
> 
> “Or perhaps a particular quality of verb-like instantiations enables it to stabilise in similarly defined rare exception. I was presented with a word I didn't previously know in a recent conversation with Jonathan Donner — gerund — a verb form which functions as a noun. Who knew?!”
> 
> 
> 
> On Wed, 16 Jun 2021 at 09:38, Henry Story <henry.story@bblfish.net> wrote:
> Hi Graham,
> 
> I think Pat would be completely ok with all the uses of ”identities” as
> described below. I am. They are really different ways of speaking of identifiers (IRIs)
> or credentials (passports), and they are compatible with two identifiers having
> the same referent.
> 
> Note that identifiers and credentials don’t  come with a list of all identity
> statements to all other identifiers in existence for that person. Ie. your passport
> does not have an extra page listing all your other identifiers. No Credential
> comes with a list of all owl:sameAs statements (the RDF name for = ) to all other
> identifiers for that person: the passport would need to constantly be updated
> and this would only work if there were really only one computer in the world, as people
> could easily otherwise coin new identifiers without the knowledge of the passport office.
> 
> 
> > On 15. Jun 2021, at 23:53, Graham Leggett <minfrin@sharp.fm> wrote:
> >
> > On 15 Jun 2021, at 18:08, Patrick J. Hayes <phayes@ihmc.org> wrote:
> >
> >> Fine, provided that y'all come up with a crisp and reasonably tight – I won't say definition, but – an account, an explanation, of what y'all mean by it, to allow readers to immediately intuit the answers to simple questions.
> >>
> >> For example, one person may have several identities, I gather. Can a person use (have? display? enact?) more than one of them at once?
> >
> > Yes. A holder of a Greek passport might be using that Greek passport to use the freedom of movement afforded to them in the EU, at the same time as that same holder of a South African passport uses internet banking to access banking services in South Africa. The Greek passport contains their name in the Greek alphabet, as well as potentially their name in the Latin alphabet. Greek surnames are gendered in Greece, but South African authorities typically assume the father’s gendered name, and so the Latin alphabet names in the two passports don’t match. The holder of these two passports are one person, but they have three names and two identities.
> 
> That is a good example. If one put that in terms of Verifiable Credentials,
> the person would have 2 Identifiers (perhaps a WebID and a DID) and two credentials.
> 
> (I wish I had taken on an Austrian passport when I was younger, then this Brexit
> chaos would have been a lot less problematic. My brother got a French one, but
> then he lives in Brazil. I know some people who have 4 passports.)
> 
> >
> >> Who or what controls which one is in use at any given moment?
> >
> > Alice does. More specifically, the Greek authorities above don’t take instructions from the South African authorities, and the South African authorities don’t take instructions from the Greek authorities. There is no control outside of what Alice wants to do.
> 
> That is the Self Sovereign Identity idea that the user holds his credentials in
> his wallet, and can should be able to choose which ones to present (with the help
> of software of course).
> 
> >> Does the person always know which one of them is in use?
> >
> > Yes, if the system they are using is well designed. Passports are separate documents that despite being standardised, are clearly distinguishable from one another.
> 
> +1
> 
> >
> >> (Is this talk of "use" even appropriate?) Can a person engage in a transaction without an identity, just being the person that they are? (Or is this impossible by definition, because interactions always involve identities rather than people?
> >
> > Yes. Alice walks into a shop, pays cash and walks out. Alice uses an Apple Music Gift Card to gain access to some music. Alice buys a one day travelcard, loads it onto a pay as you go Oyster card and uses it to travel on the London Underground. Alice buys a pay-as-you-go simcard and uses it to gain access to mobile services for a period of time.
> 
> Note that simcards don’t so much identify Alice, as they identify themselves and
> the telephone they are inserted in. In many countries they are made to identify
> a person via a law stating that people are not allowed to give the phone away
> without notifying the telecom of the change. That is you remain *responsible* for
> the sim card.
> 
> >> Or because the person /is/ one of their identities, as when we say, "Speaking for myself,…") Can more than one person have the same identity?
> >
> > A company is a group of people. The “release key” on an open source project is an identity that may represent more than one person.
> >
> >> Can things other than people have them?
> >
> > Yes. Pets can be insured. To be insured, the pets need to be identified.
> 
> That is where the criteria of identity of the referent of the identifier is important.
> So one can have for a WebID
> 
> </People/Berners-Lee/card#i> a foaf:Person .
> 
> The type foaf:Person gives us the criteria of identity of the referent.
> 
> It is given by the definition of the term at the location of the referent. That
> is illustrated here:
> 
> https://github.com/solid/authentication-panel/blob/main/proposals/HttpSignature.md#solid-use-case
> 
> but one could also have a different type of identifier that is not tied to one
> person:
> 
> <#chair> a w3c:Role .
> 
> then one could have different people chairing a Working Group for different
> periods of time, which could be expressed with some :filled relation of a role
> to time-slices of a person.
> 
> <#chair> a w3c:Role;
>     filled
>      [ startDate ”…”;
>        endDate ”…”;
>        person </People/Berners-Lee/card#i>
>      ],
>      [ startDate ”…”;
>        endDate ”…”;
>        person :Pat;
>      ].
> 
> 
> >
> >> Can an identity exist without a person, free-floating as it were? If information is given to one of them (does this even make sense?) then can the others now also access that information, or might there be things that one of them knows but the others don't? (Or is it wrong to even talk of identities knowing anything?)
> >>
> >> And so on. I genuinely have no idea what the proper answers are to questions like this.
> >
> > Identity has existed long before the internet, and if there is a plan to build anything useful the design needs to defer to real needs and experience of real life people.
> >
> > If the needs of these people aren't met, they won’t use the system.
> 
> Yes.
> 
> Someone wrote in this thread that the Internet does not have an identity
> system. Actually it does: URLs, which identify resources that
> return representations and that can be used to refer to objects via
> those representations.
> 
> We now have verifiable credentials that use the same technology.
> Certainly there are more things that need to be worked on to get
> consensus to fullfill the goals people want to put forward here.
> 
> 
> I can name quite a few, such as the need for a Web of Nations which
> I mentioned in a thread last summer
> https://co-operating.systems/2020/06/01/
> 
> But all that can build on what we have seamlessly.
> 
> Henry
> 
> 
> >
> > Regards,
> > Graham
> > —
> >
> >
> 
> Henry Story
> 
> https://co-operating.systems
> WhatsApp, Signal, Tel: +33 6 38 32 69 84‬
> Twitter: @bblfish
> 

Henry Story

https://co-operating.systems
WhatsApp, Signal, Tel: +33 6 38 32 69 84‬
Twitter: @bblfish


Received on Wednesday, 16 June 2021 13:36:28 UTC

This archive was generated by hypermail 2.4.0 : Wednesday, 16 June 2021 13:36:35 UTC