Outstanding discussion issues

[Andrew Woods]

Hello All,

For the sake of sanity, continuity, and eventual consensus/resolution, I would like to offer the following as a top-level list of open discussion topics/issues:

* Issue-1) Should F4 allow the creation of non-LDPC LDPRs?

* Issue-2) Should F4 allow triples that have non-repository subjects?

* Issue-3) Should F4 enforce trailing a '/' on LDPC URIs?

* Issue-4) Is there a bug with F4's implementation of PUT/PATCH regarding resources that are not the target of the PUT/PATCH request?

Is this an accurate listing of outstanding topics/issues? Are there other topics/issues that have been raised in the recent mega-threads?

At a minimum, I would like to arrive at concrete statements of the F4 project's position on the above, if not actual resolutions that are correct and satisfying to the majority of F4 users/stakeholders.

Thanks,

Andrew

[Esmé Cowles]

Andrew-

I agree that those are the issues I know about raised recently on the sprawling mailing list threads.

IMHO, #3 and #4 are pretty straightforward bugs that we should file in JIRA, do the research to figure out if we want to change the current behavior, etc.

-Esme

> --
> You received this message because you are subscribed to the Google Groups "Fedora Tech" group.
> To unsubscribe from this group and stop receiving emails from it, send an email to fedora-tech...@googlegroups.com.
> To post to this group, send email to fedor...@googlegroups.com.
> Visit this group at http://groups.google.com/group/fedora-tech.
> For more options, visit https://groups.google.com/d/optout.

[Andrew Woods]

Thanks, Esmé. You are probably right about Issue-3 and Issue-4.

Also, on IRC [1] today (starting at 16:09), Ben Armintor raised an additional Issue-5 from the email threads.
* Issue-5) Should RDF Resources contain triples whose subjects are not the resource; cf fcr:metadata and fcr:versions?

Regards,

Andrew

[1] http://irclogs.fcrepo.org/2015-03-16.html

[Aaron Paul Birkland]

Hi Andrew,

Comments in line

On Mon, 2015-03-16 at 14:26 -0400, Andrew Woods wrote:
> Hello All,
> For the sake of sanity, continuity, and eventual consensus/resolution,
> I would like to offer the following as a top-level list of open
> discussion topics/issues:


> * Issue-1) Should F4 allow the creation of non-LDPC LDPRs?

Also need to consider *how* (are they an object? kind of datastream?
Where would these fit into Fedora information model?)

> * Issue-2) Should F4 allow triples that have non-repository subjects?

Also need to consider objects in general, or just in LDPRs? I still
need to put strawman on wiki page related to 1&2.
https://wiki.duraspace.org/x/xJQOB

> * Issue-4) Is there a bug with F4's implementation of PUT/PATCH
> regarding resources that are not the target of the PUT/PATCH request?

My take is that the behaviour may be contrary LDP and PATCH
specs/expectation, so Esmé's suggestion to make jira tickets to
investigate would make sense.


-Aaron

[Robert Sanderson]

On Tue, Mar 17, 2015 at 7:13 AM, Aaron Paul Birkland <ap...@cornell.edu> wrote:

On Mon, 2015-03-16 at 14:26 -0400, Andrew Woods wrote:
> Hello All,
> For the sake of sanity, continuity, and eventual consensus/resolution,
> I would like to offer the following as a top-level list of open
> discussion topics/issues:

> * Issue-1) Should F4 allow the creation of non-LDPC LDPRs?
Also need to consider *how* (are they an object?  kind of datastream?
Where would these fit into Fedora information model?)

They're just like the LDPC resources, other than not allowing resources to be contained within them.

By definition, in fact, of what an LDPC is :)

 

> * Issue-2) Should F4 allow triples that have non-repository subjects?
Also need to consider objects in general, or just in LDPRs?  I still
need to put strawman on wiki page related to 1&2.
https://wiki.duraspace.org/x/xJQOB

Where else are there triples, other than in LDPRs? (given that LDPCs are LDPRs)

> * Issue-4) Is there a bug with F4's implementation of PUT/PATCH
> regarding resources that are not the target of the PUT/PATCH request?
My take is that the behaviour may be contrary LDP and PATCH
specs/expectation, so Esmé's suggestion to make jira tickets to
investigate would make sense.

It is not expected behavior, and hence other LDP implementations won't do it, it's also not at all RESTful, but I don't believe the PUT behavior is forbidden as LDP does not make requirements on processing the entity-body of requests.  PATCH on the other hand is specified here:  http://www.w3.org/TR/ldpatch/

Rob

--

Rob Sanderson

Information Standards Advocate

Digital Library Systems and Services

Stanford, CA 94305

[aj...@virginia.edu]

On Mar 17, 2015, at 11:02 AM, Robert Sanderson <azar...@gmail.com> wrote:
> On Tue, Mar 17, 2015 at 7:13 AM, Aaron Paul Birkland <ap...@cornell.edu> wrote:
> On Mon, 2015-03-16 at 14:26 -0400, Andrew Woods wrote:
> > * Issue-1) Should F4 allow the creation of non-LDPC LDPRs?
> Also need to consider *how* (are they an object? kind of datastream?
> Where would these fit into Fedora information model?)
>
> They're just like the LDPC resources, other than not allowing resources to be contained within them. By definition, in fact, of what an LDPC is :)

I appreciate the point, Rob, but that does nothing to answer Aaron's questions (which are also my questions). Fedora's model _is not_ the LDP model, so answering the question using LDP constructs is no answer. We're trying to find an answer using Fedora constructs (objects, datastreams, and the relationships amongst them).

> > * Issue-2) Should F4 allow triples that have non-repository subjects?
> Also need to consider objects in general, or just in LDPRs? I still need to put strawman on wiki page related to 1&2. https://wiki.duraspace.org/x/xJQOB
>
> Where else are there triples, other than in LDPRs? (given that LDPCs are LDPRs)

I think Aaron here meant specifically non-LDPC LDPRs. Perhaps there is a precise term for those guys?

---
A. Soroka
The University of Virginia Library

[Aaron Paul Birkland]

Hi all,

I added a proposal in the wiki, it's at he bottom:
https://wiki.duraspace.org/x/xJQOB

It's called "LDP-RS as datastreams"

If I get time later today, I'll add a section called something like 'use
cases' or 'motivation', if that is unclear to anybody

-Aaron

[Scott Prater]

Aaron,

I renamed the page, to make it more general, rather than focusing on
external datastreams.

Could you include some sample RDF, something that you would expect to
get back from Fedora if you requested a non-LDPC LDP-RS object, to
illustrate the points in your proposal?

-- Scott

--
Scott Prater
Shared Development Group
General Library System
University of Wisconsin - Madison

[Benjamin Armintor]

I think the utility of LDPRs is obvious, but I think that building them under the datastream concept is weird. Is it just an end-around on server-managed triples? 

There's already fiddly bits about treating RDF data as non-RDF data. Adding more fiddling here to support PATCH and have a RDF-not-RDF act like RDF again... Doesn't this proposal boil down to:

1. +1 for LDPR

2. Proposed: Excluding server-managed triples

... and if so, can we discuss it in those terms? Except for PATCH, the rest of the Datastream stuff is already available (unless I'm missing something).

- Ben

--
You received this message because you are subscribed to the Google Groups "Fedora Tech" group.

To unsubscribe from this group and stop receiving emails from it, send an email to fedora-tech+unsubscribe@googlegroups.com.

[aj...@virginia.edu]

I'm grateful to Aaron for providing the beginnings of an explanation of how these guys fit into the Fedora model, without which I have no intention of changing my vote on the original proposal. Ben-- I appreciate your desire to move entirely over to LDP and abandon the Fedora model, but I don't share it. As far as the "adjustable" exclusion of server-managed triples, I understood that to be pretty non-controversial all around.

There is another, more crucial point in Aaron's strawman:

> Allow the LDP-RS-NCs created by Fedora to contain arbitrary RDF
> • No requirements on allowable subjects, predicates, or objects in the triples
> • No constraints on constructs such as blank nodes

I couldn't possibly be more opposed to this idea, for reasons that have been well-discussed on this list very recently, particularly by Rob Sanderson.

---
A. Soroka
The University of Virginia Library

> To unsubscribe from this group and stop receiving emails from it, send an email to fedora-tech...@googlegroups.com.

> To post to this group, send email to fedor...@googlegroups.com.
> Visit this group at http://groups.google.com/group/fedora-tech.
> For more options, visit https://groups.google.com/d/optout.
>
>

> --
> You received this message because you are subscribed to the Google Groups "Fedora Tech" group.

> To unsubscribe from this group and stop receiving emails from it, send an email to fedora-tech...@googlegroups.com.

[Neil Jefferies]

My main use case would be for the LDP-RS-NC to contain triples that
relate to itself. That is, it is essentially a self-contained,
implicitly named graph or an object whose contents are just RDF.

For unconstrained triples - wouldn't this be better as a container that
contains "unmanaged-or-externally-managed" RDF datastreams that are
basically treated like other non-RDF datastreams?

N

[Benjamin Armintor]

My desires about the Fedora model aside, I think it is worth considering how, or if, this construct is like a Datastream, or whether it is more like the other Fedora type, which is represented in Fedora 4 as an RDFSource.

[aj...@virginia.edu]

With that I certainly agree, if I correctly understand you to mean by "the other Fedora type" what we have for fifteen years referred to as a Fedora object. It seems to me that there are two options that have been discussed:

1) These constructs appear as "pointlike objects": objects with no datastreams and no contained objects, only properties.

2) They appear as special datastreams. (This is what Aaron is suggesting.)

I'm inclined to the former, for two reasons: one, because it allows us to treat them as limited or restricted versions of something that is already present in the model (as opposed to the latter option, which treats them as datastreams with new abilities), and two, because none of the use cases I have yet heard actually use them like datastreams. E.g. Ben's example of entities that represent vocabulary terms in a repository clearly treats them as things with their own ontological identity-- a clear marker of being an object in the Fedora model.

---
A. Soroka
The University of Virginia Library

[Aaron Paul Birkland]

On Wed, 2015-03-18 at 11:59 -0400, Benjamin Armintor wrote:
> I think the utility of LDPRs is obvious, but I think that building
> them under the datastream concept is weird.

Why weird? Think of it this way. Right now, Fedora allows one kind of
LDPR: a LDP-NR. The conceptual mapping of "object" to LDPC and
"datastream" to LDPR is already present. The proposal builds on that:
map the other kind of LDPR resource (an LDP-RS) to a datastream as well.

I think this point is easy to get across in documentation. LDP
container -> Fedora object. LDP non-container resource -> Fedora
datastream. The interaction model defined by LDP for RDF resources is
REALLY USEFUL, and the difference in interaction modes between RDF and
non-RDF datastreams can be explained by concepts vetted by the LDP
community.

> Is it just an end-around on server-managed triples?

No - that's one part. The other part is "no specific RDF
constraints" (like restrictions on subject). This is what makes
"datastream" attractive to me. It's just a blob BUT with RDF semantics
and very useful interaction mode promised by LDP.

[Aaron Paul Birkland]

On Wed, 2015-03-18 at 12:07 -0400, aj...@virginia.edu wrote:
> I'm grateful to Aaron for providing the beginnings of an explanation of how these guys fit into the Fedora model, without which I have no intention of changing my vote on the original proposal. Ben-- I appreciate your desire to move entirely over to LDP and abandon the Fedora model, but I don't share it. As far as the "adjustable" exclusion of server-managed triples, I understood that to be pretty non-controversial all around.
>
> There is another, more crucial point in Aaron's strawman:
>
> > Allow the LDP-RS-NCs created by Fedora to contain arbitrary RDF
> > • No requirements on allowable subjects, predicates, or objects in the triples
> > • No constraints on constructs such as blank nodes
>
> I couldn't possibly be more opposed to this idea, for reasons that have been well-discussed on this list very recently, particularly by Rob Sanderson.

Right, and I couldn't possibly be more opposed to the idea of Fedora
telling me what I should or should not to be putting in my datastreams.

-Aaron

[Aaron Paul Birkland]

On Wed, 2015-03-18 at 16:23 +0000, Neil Jefferies wrote:
>
> For unconstrained triples - wouldn't this be better as a container that
> contains "unmanaged-or-externally-managed" RDF datastreams that are
> basically treated like other non-RDF datastreams?

Right. The issue is that LDP defines a very useful interaction mode for
RDF content that goes beyond binary/blob. I feel that if Fedora
embraces that interaction mode for RDF digital assets (as datastreams),
that would be a significant win.

-Aaron

[aj...@virginia.edu]

Aaron, unless your proposal goes through in exactly its current form, what you're talking about aren't datastreams at all (and as I mentioned in my last message, I'm not sure that they should be).

In some ways, we're talking about the difference between thinking about these triples as miscellaneous RDF data (which is how I understand you to be thinking of them) and thinking of them as representing the properties of things in the repository (which is how I am thinking of them).

If they are just RDF data, then you are certainly right that Fedora has no business constraining them. But if they are just RDF data, stick them in an opaque datastream and Fedora won't constrain them. If they are, instead, properties of resources defined as Fedora resources, then Fedora has a legitimate need to constrain them.

This seems to me perhaps a little more evidence that what we are talking about "pointlike objects" and not datastreams at all.

---
A. Soroka
The University of Virginia Library

[Aaron Paul Birkland]

Right - my perspective is that a non-container LDP-RS would just be RDF
data, and that if Fedora added the interaction mode specified by LDP for
these datastreams, it would be incredibly useful, convenient, and maps
very well to LDP constructs.

I don't quite fully understand the 'pointlike object' concept. If it's
truly being considered as an option, could you or someone else write
something on the wiki about it? What benefit do they have that
existing Fedora objects/containers don't have? Is it merely the
constraint that they cannot have datastreams that is of value?

My own feeling at this point (unless I'm missing something) is that
Fedora gains a lot less through pointlike objects than through allowing
the LDP RDF interaction model for RDF-containing datastreams.

-Aaron

[aj...@virginia.edu]

You're right to push me to the wiki, but just to be clear, the "pointlike object" for me is just an object without datastreams or contained objects, and the crucial point (pun intended!) is that since objects already support (most of) the LDP interaction model, there is no need to add some new idea of a special datastream-type with extra functionality.

Datastreams would remain as they always have been for Fedora-- opaque bitstreams in the repository with some properties.

---
A. Soroka
The University of Virginia Library

[Andrew Woods]

Hello Adam,
Your conceptualization of LDP-RS-NCs as "pointlike objects" fits into the existing F4 model fairly well, although the suggestion may be less contentious if the name were "pointlike resources". Regarding Aaron's "Example 2" [1], do you see a "pointlike object" approach requiring updates to the proposal beyond removing the line-item (2a)?

-----
(2a) No requirements on allowable subjects, predicates, or objects in the triples

-----

Also, is there any reason why you may not expect an LDP-RS-NC to abide by an LDPR interaction model?

Hello Aaron,
I do not exactly understand what you are driving towards with the proposal's [2] point (2).

-----
(2) Fedora may describe these resources using slightly different metadata in their corresponding 'fcr:metadata' resources.  For example, LDP-RS-NCs may have a different mixin type than LDP-NRs

-----

Andrew
[1] https://wiki.duraspace.org/display/FF/Design+-+Modeling+for+non-LDPC+objects#Design-Modelingfornon-LDPCobjects-Example2:LDP-RSasdatastreams

[2] https://wiki.duraspace.org/display/FF/Design+-+Modeling+for+non-LDPC+objects#Design-Modelingfornon-LDPCobjects-Proposal

[aj...@virginia.edu]

Andrew-- I've gone ahead as Aaron wisely advised and put a small proposal on our design/discussion page:

https://wiki.duraspace.org/display/FF/Design+-+Modeling+for+non-LDPC+objects#Design-Modelingfornon-LDPCobjects-Proposal:%22Pointlikeobjects%22

I have no objection to calling these guys whatever you like, so long as we understand them clearly to be Fedora objects and not datastreams-- that is the _point_ of the proposal! (See what I did there?)

Of the four items Aaron addresses, the only ones in common between our proposals are the first and last. I object entirely to the second ("Allow the LDP-RS-NCs created by Fedora to contain arbitrary RDF") and the third ("Allow the LDP-RS-NCs created to by Fedora to be 'pure' in that they are logically unmodified from client requests") seems to me not to belong in this discussion at all, as I said earlier.

With that, I'll suggest that we take this discussion to comments and edits on that page, to avoid pushing too many threads at once through the mailing list.

---
A. Soroka
The University of Virginia Library

[Andrew Woods]

Hello All,
It would appear that this conversation has settled to a stop, after having taken an almost immediate deep-dive into the initial issues 1 & 2 below.

* Issue-1) Should F4 allow the creation of non-LDPC LDPRs?

* Issue-2) Should F4 allow triples that have non-repository subjects?

* Issue-3) Should F4 enforce trailing a '/' on LDPC URIs?

* Issue-4) Is there a bug with F4's implementation of PUT/PATCH regarding resources that are not the target of the PUT/PATCH request?

* Issue-5) Should RDF Resources contain triples whose subjects are not the resource; cf fcr:metadata and fcr:versions?

Acknowledging that the following next-steps can not address all interests raised, they do move Fedora forward in a constructive direction based on the inputs and discussion on this and related threads. Here is a status update on the next-step actions:

+ Issue-1) Yes.

Related tickets:
https://jira.duraspace.org/browse/FCREPO-1396
https://jira.duraspace.org/browse/FCREPO-1470

+ Issue-2) No, RDF triples that go into the repository (unless packaged as an opaque datastream) should have the repository target resource as their subject.

Related ticket:
https://jira.duraspace.org/browse/FCREPO-1411

+ Issue-3) Not for now.

Related ticket (on hold):
https://jira.duraspace.org/browse/FCREPO-1410

+ Issue-4) Yes, Fedora should return the appropriate error responses and code when it prevents the updating of resources that are not the target of a PUT/PATCH request.

Related tickets:
https://jira.duraspace.org/browse/FCREPO-1417
https://jira.duraspace.org/browse/FCREPO-1418

+ Issue-5) No, RDF triples that come out of the repository should have the requested resource as their subject.

Related ticket:
https://jira.duraspace.org/browse/FCREPO-1474

Regards,
Andrew