Alexandre Monnin (Paris 1, IRI, INRIA) , Nicolas Delaforge (INRIA), Fabien Gandon (INRIA) April 17, PhiloWeb 2012, WWW2012, Lyon, France

From linked documentary resources to

linked computational resources

Goal.

We wish to demonstrate that it is possible to

account for the putative transition between a

Web of document towards a Web of

applications strictly from an architectural point

of view.

Web Architecture:

Three (?) main concepts

URIs HTTP

HTML

/RDF REST

Assumption: Blogic is

right 1/1.

"<can-of-worms> Note, this is about what URIs refer to

when used as logical names, not what they "identify" when

used by HTTP. These are two quite distinct ideas. Typically

(not always) a URI identifies some (source of) data about

what it refers to. </can-of-worms>"

(Pat Hayes)

Assumption: Blogic is

right 2/2.

Two visions. Blogic and RDF as it is (aka the Semantic

Web?)

- The Web comes first

- We deal with HTTP URIs

- Resources vs representations

- RDF

- We deal with URIs as proper names

- Meaning of URIs

Two approaches, not two

Webs

- The RDF is one take on the Web, a model,

not as complex as reality.

- "Death by layers": but look, we got those

layers already so we should think about

them! How they relate to the Web (theory of

assemblages).

Cake!

Resources (my - AM -

take)

- Resources aren't things out there : you don't

need to previously check a thing exists in the

physical or scientific sense of the word to

identify a resource.

- Cannot be accessed, we all know that

("shadows", if that is not a means without an

end what is it ! Anything at all...).

- The resource is what bears on the

representations, what explains why there were

picked up.

In REST (birthplace)

Three things:

1. Resource

2. The states of a resource

3. The representational states of the resource

1. Rule

2. Application of the rule

3. Representation of that result

Resources are anything,

but also...

" (...) the semantics of what an author identfies".

(Fielding and Taylor 2002)

Just an author ?

In Webarch this idea seems to come from Kripke's idea of

baptism. There would lot to say but let's not discus this

now.

Let's rather find out if what we need is such a model of

authorship.

Micro, meso, macro.

The micro level focuses on the resource itself and its inner

mechanisms.

The meso level is about relations and interactions between

computational resources.

The macro level highlights the causal relations between an

editorial policy of a publisher and the way he manages

his web resources

Micro. More clients.

Many more devices are becoming clients.

"Web servers were originally designed to propose "filesystem like" remote

services. Since the common gateway interfaces (CGI) their structure have

become increasingly complex. Nowadays, servers are able to negotiate with

clients to adjust the response so that most of the content is generated on the

fly. Any Web server is also compatible with at least one programing language

that can trigger the processing of very sophisticated tasks that sometimes

involve other remote services."

Micro. Break the causal

relation.

"One of the defined rationale behind documentary

resources is that people have tried to preserve the

causal relation between a reference and an

informational content, because it was constitutive of all

our "real world" documentary reference systems. The

evolution from documentary resource to computational

resource made more obvious that this artificially

preserved causal relation had been broken."

Micro. Conclusion.

The documentary location has been replaced by a locus of

computation, a space of invocation.

CGI and REST have turn URL into RPC passing

parameters to scripts or web services. Now everything is

(and has always been in a sense) URI which are

identifying protean resources that can turn themselves in

any format required by the client. Such are the

computational resources.

Micro to macro.

As said before, a resource is a formal translation of a

publishing rule but these rules can change, the

implementation can evolve to match a new technological

context, a bug can be fixed, a database can be updated

with fresh data. There are many reasons for Web

representations to change and that is the true

communication power of the Web, an editor can instantly

adapt the whole editorial chain synchronously with any

informational/technological activity.

Meso

Through HTTP, any computational resource is likely to

refer to other resources or to communicate with them

E.g.:

• Web services composition and orchestration

• Web Data transformations and Mashup

More than ever, resources are related to each other and

can be nested to create original compositions. Thus,

qualifying the Web as an hypertext seems a little bit

outdated so we would rather talk about hyperprocess.

Macro. Many more roles!

1/2

This seems to become clearer everyday.

The authorship model was maybe related to the

"documentary Web". A Web of addresses

and static documents. Not that it ever was

like that but it is the way it looked like and

was thought of and used.

Macro. Many more roles

2/2

Too many to mention but a few:

• URI "minter"

• Resource definer?

• Resource publisher

• Service provider

• Information architects

• etc.

Macro. May be one person

Macro. More often not...

Macro. Computational

commitment.

On the other hand, it is more and more difficult for

publishers to ensure a good quality of service throughout

the processing chain. The technological stack and the

processes involved in publishing a resource have become

so complex and so distributed that it is becoming harder

and harder to ensure a strict editorial commitment because

as the Web grows in diversity, this commitment has turned

into a computational one.

Macro. Many more rules

(1/2).

The resource is not the only rule : how individual resources

are distinguished from one another depends on a

publishing commitments. Other rules, more or less

implicit.

In other words, Web resources are often published as part

of bigger resource sets, that have in common to be named

and managed by the same publisher.

Macro. Many more rules

(2/2).

We consider that an editorial policy can be summarized as a structured

rule set. Some of these rules are generic, some others are specific

and can inherit or be related to broader ones. From this, we assert

that any Web resource formally expresses one of these publishing

rules. In other words, a Web resource is situated at the intersection

of a number of publishing rules.

A URI then gives access to a representational state that is the result of

this intersection and its closure, while it only identifies the most

specific rule involved in generating the aforementioned

representational state.

Macro: editorial

commitment.

From the societal point of view, content publishers whose

main activity was to produce content and to guarantee the

quality of information now have to deal with various new

constraints owing to the specificity of the medium.

Conclusion.

The architecture of the Web of data and the models of the Semantic

Web may provide a way to match the diversity of online resources by

means of a framework of metadata designed to annotate Web

resources and exploit the semantics of their schemas to process them

intelligently.

Metadata and their schemas could be the keystone of the new

resource-centric Web applications, their integration and interoperability.

It is conceivable that tomorrow, he who controls metadata on the Web,

controls Web resources, and through them a lot of things.

Thanks!