File Summaries vs. Quad Component Summaries

[simonstey]

We say:

In the proposed framework, data pods expose a separate summary for each file,

=> 1 summary per file which would be in line with what's shown in this figure:

however, we continue with stating that:

[..]. We assume that pods expose summaries that are created according to the algorithm presented in Listing 1. In this algorithm, a file summary is created for each quad component, where we iterate over all the file’s quads, and the access token (aka a key) that are applicable for each quad.

=> 1 (file) summary for each quad component, i.e. 4 summaries per file?

Considering our algorithm for creating privpres summaries -> https://github.com/rubensworks/article-adecentweb2020-privacy-decentralized/blob/520dec5c5ad42bfb5cd56561424e73ba4fe31ea6/content/code/summarization-algorithm.txt#L1-L14

I reckon it's:

1. there's only 1 summary for each resource/file
2. a file's summary consists of 4 components/a dictionary containing its subjects/props/objs/graphs
3. if we talk about summaries we refer to those "top level" summaries, i.e., the ones of which each file has exactly 1 one of and that are exposed by the pod and used by aggregators

right?

[rubensworks]

1. there's only 1 summary for each resource/file

👍

2. a file's summary consists of 4 components/a dictionary containing its subjects/props/objs/graphs

Sure, "components" sound like a good term to me.

3. if we talk about summaries we refer to those "top level" summaries, i.e., the ones of which each file has exactly 1 one of and that are exposed by the pod and used by aggregators

Exactly.

[simonstey]

could we call those 4 components a dictionary? kinda reminiscent of HDT's concept of a dictionary:

The dictionary is divided into sections depending on whether the term plays subject, predicate, or object roles.

but extending it with a section for graphs

[rubensworks]

Dictionaires are used to provide a concrete mapping between things, like string -> int. Since that is not exactly what we do here, I would not use that term, to avoid confusion.

[simonstey]

Dictionaires are used to provide a concrete mapping between things, like string -> int. Since that is not exactly what we do here, I would not use that term, to avoid confusion.

true.. anyway I think we should rephrase things like

OUTPUT: 
     summary for each quad component: Σ.subject, Σ.predicate, Σ.object, Σ.graph 

to something along the lines of:

OUTPUT: 
    a summary Σ consisting of four "parts/components" (?), one for each quad component: Σ.subject, Σ.predicate, Σ.object, Σ.graph 

otherwise one might expect to get 4 individual summaries, while it's actually just 1 summary having 4 components/parts/...

[rubensworks]

Sure, I agree with that :-)