Re: References and Modularity

Marcos Caceres [2013-06-16T04:47]:
> Call them what you want, doesn't make a difference :) same crap, different shades. 

There is a semantic difference. And btw I'm not contradicting the basis of your idea, just the choice of words.

    Your idea: Using a model where 
               the identifier (link) to the specification has to return 
               the always most recent version of a representation.
    What you are argueing against:
               Using a model where there are stages of maturity 
               for specific representations of a technology.

It is not related to the identifier scheme model (dated uri, hash uri, md5 uri, integer uri, etc.).  
I guess I could draw it.

* representation     (content)
* versioning         (identifiers scheme model)
* maturity/milestone (chosen process model)


In your idea something that is interesting and should be articulated to understand the consequences:

* To be sure that implementers have access to the always most recent version, is it possible to remove any references to a versioning systems of identifiers. I would say for a metaphor, writing on the sand at the same place. There is no archive of the historicity. It means no dvcs. I guess the consequence would be that people would create their own archiving systems (recreating memories) depending on their contexts. Something ala http://web.archive.org/ but that could be fine it would be their responsibility. 


* Or is it just a matter of metadata in the representation allowing the possibility to go to the always most recent version, aka the start of the document with links to different type of representations. The introduction in the last few years of the "Editor's draft" link is something along that. What *you seem* to say is  that the specification shortname uri (a specific identifier) should be the Editor's draft shortname uri. 


So let's say we have that model in place. Stages of maturity don't exist anymore by design, because there is no specific version to point to it. So there is no LC, no CR, no nothing. There is just what you called a "living document". And each reference is toward a representation which is always the last one. It also means we remove the notion of conformance, because conformance is a concept which is about being similar to a reference. You cannot be similar to something which is moving all the time, except if you are that thing (identity).

An implementer would have a software which is always an incomplete version of what should be done (except if the representation stops evolving but in this case it is not a living document, it is a dead one). It could be fine. It would not work for lawyers at all and patent policies which are systems which rely on specific versions (there is a notion of history in patents). I wonder if it would remove a hook to the patent system at all.  

Btw the current whatwg specifications are not exactly into this model (writing on the sand), there are using a dated version under the form of hash-uri with the link to the "Version History:" see for example http://url.spec.whatwg.org/ 
The whatwg specifications have no maturity/milestones model, aka specific versions with an identifier attached to a precise meaning in a process.


# Issue: 
    Why is there a date in the content of a living document? For example http://url.spec.whatwg.org/ 
    Somehow it is no sense. It recreates history and then versioning.


It might sound meta, or philosophical but it is exactly what we are discussing. I have for myself not yet make my mind. I want to explore and understand the ideas. 


-- 
Karl Dubost
http://www.la-grange.net/karl/

Received on Sunday, 16 June 2013 09:57:44 UTC