Re: ldp-ISSUE-48 (profiles): Profile mechanism is Needed [Linked Data Platform core]

On 31 Jan 2013, at 19:37, "Wilde, Erik" <Erik.Wilde@emc.com> wrote:

> hello henry.
> 
> On 2013-01-31 19:18 , "Linked Data Platform (LDP) Working Group Issue
> Tracker" <sysbot+tracker@w3.org> wrote:
>> ldp-ISSUE-48 (profiles): Profile mechanism is Needed [Linked Data
>> Platform core]
>> It has been argued again and again by Erik Wilde on the mailing list [1]
>> that LDP needs a way to be able to specify what types of triples must
>> appear in a graph for a specific resource. Others have argued that one
>> needs this to specify what type of entity should be POSTed to a server.
>> The notion of profiles has come up as name for specifying this [2].
> 
> this is confusing two issues. one issue is that resources that are POSTed
> must be self-describing, so that for example a LDP service can understand
> the difference between a container being POSTed and a member being POSTed.
> for this, it is sufficient if the protocol specifies preconditions that
> have to be satisfied by the POSTed RDF, and then the service can dispatch
> after validation based on what it found in the POSTed resource. for this,
> no profiles are required, all we need is to make the representations
> self-describing.

The issue as I explain below is how to define these preconditions.

> 
> the other issue that it would be good if LDP had a presence in the uniform
> interface of the web, so that it can be advertised and detected in
> scenarios where discovery and dispatch is based on media types. it is
> unlikely that LDP will define its own media type. it could be argued that
> LDP could still be made visible by using profile media type parameters,
> but even that would require that the various RDF syntaxes that might be
> used to expose LDP all start supporting a profile parameter. this might
> also be a tough sell to the larger RDF community. if we use generic RDF
> media types, LDP will be invisible on the media type level.

The issue of media types or other is a side issue. 

Those are just ways of signaling information. You can send
this signal with media types, which is clumsy 

  Content-Type: text/rdf; type="http://www.w3.org/ns/ldp#EntryProfile"

or with an elegant link type 

  Link: <http://www.w3.org/ns/ldp#EntryProfile>; rel="profile"

But neverymind. That is a minor issue. Either of those could
easily be chosen.

The issue here, is that you are claiming that one needs to define 
somehow what appears in the graph, so that user agents can 
know what to expect when the fetch a graph or send one. 
We therefore need  a language  to specify these constraints.  We 
cannot do it with DTDs because RDF does not have a precise 
syntax. So XSLTs and XQueries would also not work.

In short: if you are right this profile language that cuts across
formats has to be defined. 

> 
> cheers,
> 
> dret.
> 
> 

Social Web Architect
http://bblfish.net/