[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index] [List Home]
[higgins-dev] Relative and absolute UDIs

Inline

 


From: Anthony Nadalin [mailto:drsecure@xxxxxxxxxx]
Sent: Wednesday, April 23, 2008 12:29 AM
To: Paul Trevithick
Cc: 'Higgins (Trust Framework) Project developer discussions'
Subject: RE: [higgins-dev] I've updated the HGG wiki and PPT

 

My point is that there is no global graph as you may only have relative identifiers (and that is perfectly OK),

I’m trying to understand what you’re saying. Today in the Higgins model all Entities live within Contexts and all of these Contexts are identified by ContextIds. Today we do allow ContextIds to be “local” –that is, NOT discoverable globally. Here’s how. The creator of a Context can choose to describe it in an XRDS file that is behind a firewall or even in a file on the local machine. Entities living within such a “local” Context are thus NOT part of the global graph. I do admit that the Higgins wiki doc has a “global” bias, whereas I realize in writing this email that many Context developer/deployers may choose to keep their Contexts and the Entities they contain off the global grid, so to speak.  

also there are many other types of universal identifiers besides URIs. Lets take the case of the phone number the context ma be country code (+1) and the entity may be the area code and local number (512-222-2222) thus I have a universal identifier that is not a URI,

If and only if an Entity needs to be addressable from OUTSIDE its local Context (e.g. if the Entity needs to be a part of the global graph, or even if it needs to be addressable in a graph that spans Contexts but is entirely behind a firewall) then it needs an EntityId. So if the Entity doesn’t need to be addressable then it doesn’t need an EntityId at all. At the Provo F2F (for Higgins 1.1) we, instigated by you, introduced the new idea that some Entities do not have EntityIds. We call these BlankEntities. So we have a way of dealing with Entities that are just like elements in a relational database —not directly addressable at all.

OTOH, if an Entity needs to be addressable from OUTSIDE of its local Context, then it does need an EntityId. Its containing Context’s ContextId is the prefix of the EntityId and a string (e.g. a telephone number) forms the rest of the EntityId. If an Entity wishes to use a “global” (e.g. a +1…) telephone number as the relative part of the identifier, then that’s okay too. But the ContextId prefix is still required. And since containing ContextId must be a UDI URI (e.g.  http://foo.com ) then the contained EntityIds would look like this: http://foo.com#+1-512-222-2222 , etc. (modulo necessary character encoding/escaping)

also not sure one needs a universal identifier as I may only want to reference within a context so I only need the 512-222-2222).

The “local” Entity (that you call 512-222-2222) would have an EntityId of http://foo.com#512-222-2222 where http://foo.com (as above) is the ContextId of the containing Context.

BTW, in the Higgins 1.1 IdAS implementation of the HGG model, IEntity interface perhaps should have a “getLocalId()” method that just returns the relative/local EntityId (512-222-2222 or +1-512-222-2222 or whatever) in addition to the “getEntityId()” method that returns the absolute EntityId.

So not in favor of this UDI as you describe, and we should discuss, these changes just seem to pop into place w/o any discussions.

UDI is just a new name for a concept that has been in Higgins from 2003. That’s why I’m saying we have a terminological issue here. The UDI resolution is already implemented in Higgins 1.0 in the IdAS Context Registry (see [1]). UDI is analogous to the role that URLs play to knit together the web. Try to imagine the web (HTTP and HTML) without URIs.

[1] http://wiki.eclipse.org/IdAS_Registry

Anthony Nadalin | Work 512.838.0085 | Cell 512.289.4122

Inactive hide details for "Paul Trevithick" ---04/22/2008 05:38:56 PM---I think we just have a terminological issue here. I agr"Paul Trevithick" ---04/22/2008 05:38:56 PM---I think we just have a terminological issue here. I agree that a telephone number string can certainly be used as the relative


From:


"Paul Trevithick" <paul@xxxxxxxxxxxxxxxxx>


To:


Anthony Nadalin/Austin/IBM@IBMUS


Cc:


"'Higgins \(Trust Framework\) Project developer discussions'" <higgins-dev@xxxxxxxxxxx>


Date:


04/22/2008 05:38 PM


Subject:


RE: [higgins-dev] I've updated the HGG wiki and PPT





I think we just have a terminological issue here. I agree that a telephone number string can certainly be used as the relative portion of a UDI (e.g. the phone number in this UDI: https://example.com/contexts/context100.xrds#617-555-1212 that uses an HTTP URI to get to its XRDS document) to identify an Entity. And when it is prepended with its containing Context’s id (aka ContextId) we now have that Entity’s absolute UDI. [It is also true that Entities can have any number of additional attributes any of which can be used for queries and thus act as identifiers to distinguish one Entity from others.]

Without universal data identifiers (UDIs) (formerly called Higgins identifiers) there is no “global graph” (or “data web” as some call this kind of thing, or the “linked data” web that the semweb folks call it) at all. By requiring that every Entity have a UDI that persistently identifies it, you can create Entity-to-Entity relationships, correlations, federation, etc.

See also http://en.wikipedia.org/wiki/Linked_Data for closely related ideas.


From: higgins-dev-bounces@xxxxxxxxxxx [mailto:higgins-dev-bounces@xxxxxxxxxxx] On Behalf Of Anthony Nadalin
Sent:
Tuesday, April 22, 2008 4:07 PM
To:
Higgins (Trust Framework) Project developer discussions
Cc:
higgins-dev-bounces@xxxxxxxxxxx; 'Higgins (Trust Framework) Project developer discussions'
Subject:
Re: [higgins-dev] I've updated the HGG wiki and PPT

There should be no identifier defined, this is just an attribute and the attribute does not have to be a URI or a XRI, this is far too constraining as it can be a telephone number, etc. Note that an identifier is not the same as the ContextID or EntityID.

Anthony Nadalin | Work 512.838.0085 | Cell 512.289.4122

Inactive hide details for "Paul Trevithick" ---04/22/2008 02:13:43 PM---Jim,"Paul Trevithick" ---04/22/2008 02:13:43 PM---Jim,


From:


"Paul Trevithick" <paul@xxxxxxxxxxxxxxxxx>


To:


"'Jim Sermersheim'" <jimse@xxxxxxxxxx>


Cc:


"'Higgins \(Trust Framework\) Project developer discussions'" <higgins-dev@xxxxxxxxxxx>


Date:


04/22/2008 02:13 PM


Subject:


[higgins-dev] I've updated the HGG wiki and PPT






Jim,


The wiki pages rooted at [1] have now been updated per Higgins 1.1 in three ways. First, the sub-pages all have (or should have) a “Version” section at the top. Second I have eliminated the 1.0 concept of Entity and changed zillions of places where Node was to now read Entity. Finally, I’ve update a few of them per our decision to simply use Entities as complex values. I have also edited the PPT here [2] per the latest HGG 1.1 model.


I also renamed “Higgins Identifier” to “UDI” and expanded its coverage from HTTP URIs and XRIs to also include support for the semweb community’s “Linked Data” HTTP URIs. We haven’t yet implemented support for Linked Data URIs but I think that doing so would allow Higgins IdAS to be used as a gateway to at least some of the emerging Linked Data data sources. Parity will be contributing resources to implement this stuff. It will help the IdAS/Higgins cause WRT both the partially overlapping semweb and the dataportability crowds.


-Paul


[1]
http://wiki.eclipse.org/Higgins_Global_Graph_1.1
[2]
http://dev.eclipse.org/viewsvn/index.cgi/org.eclipse.higgins/trunk/doc/org.eclipse.higgins.doc/Higgins-Data-Model-Intro.ppt?root=Technology_SVN&view=co
_______________________________________________
higgins-dev mailing list
higgins-dev@xxxxxxxxxxx
https://dev.eclipse.org/mailman/listinfo/higgins-dev

GIF image

GIF image

GIF image

GIF image

GIF image