Re: web intents agent !== user agent

On Tue, Jun 5, 2012 at 2:14 AM, Jean Claude Dufourd
<jean-claude.dufourd@telecom-paristech.fr> wrote:
> Thanks Claes, that discussion was about the core of my question. I am not sure about the cloud part, but the separation from UA is key. So to answer Greg's question on defining the WIA, The WIA is the part that would constitute the separate entity to which the UA would delegate all of the intent-related processing.
>
> In my arguments, I touched on multiple applications for such a separation between WIA and UA, such as user perspective, hierarchy of circles and security, but they are all enabled by the ability to separate the WIA and the UA.

What I hear you saying is something like this: there should be a
standard way for the UA and other intent-interested apps on a device
to register with some kind of system-wide broker such that they can
exchange intents.

That is a good idea, but I don't think it is in scope for the web
intents spec. This spec is about how web content invokes intents,
registers for them, and has them delivered. So it appropriately
governs the UA in that the UA is the platform for web content. It's
out of scope for the web intents spec to govern anything else -- UA
developers have no control over OS-level features.

Or do you mean more that web content should be able to replace the
picker itself? That seems like a good idea for a UA extension API, but
I don't think it's a good idea to add to the spec, at least at this
point. Since the picker mechanics are under the control of the UA,
there will be different extension mechanism required for different UA
implementations.

Something that is available to web content under the spec is explicit
intents. That is, an app can allow users to register services, and
then provide picker-like mechanics for them, all internally. This is
only in the context of that app, of course, but that's to be expected,
right?


> My main point is that the spec is written in such a way that WIA is not even defined separately from the UA. Web Intents is a fairly complex spec, and despite being in the field and having followed the discussions and having implemented related stuff, I do not trust myself to extricate when UA in the spec means WIA and when it's only UA, for the purpose of implementing a separate WIA system.
> Actually, in some places, UA is mentioned where if separated, there would be a communication between WIA and UA, one way or the other. These are the most difficult spots, realistically.
> So this de facto merge is:
> - a barrier to innovation by most by increasing the cost of separation,
> - creating a large risk to interoperability by leaving the separation for implementors to figure out, and ensuring that no two will get it exactly the same, resulting in incompatible systems.
>
> If, as I believe, the separability of WIA and UA is important, it has to be explicitly reflected in the spec.
> Best regards,
> JC
>
> On Jun 5, 2012, at 0:17, "Nilsson, Claes1" <Claes1.Nilsson@sonymobile.com> wrote:
>
>> Hi Greg,
>>
>> In Shenzhen we touched the issue that Jean-Claude addresses. According to my notes the question was if we could have a Service registry in a cloud server instead of locally in the browser. That would allow several User Agents belonging to a user share the registered Services. My impression of the conclusion was that there is nothing in the Web Intents model that prevents this but it is currently not clearly specified.
>>
>> This is not the whole issue that Jean-Claude addresses but a part of it.
>>
>> Best regards
>> Claes
>>
>> -----Original Message-----
>> From: Greg Billock [mailto:gbillock@google.com]
>> Sent: den 4 juni 2012 19:39
>> To: Jean-Claude Dufourd
>> Cc: public-web-intents@w3.org
>> Subject: Re: web intents agent !== user agent
>>
>> Can you say more about what you mean by the "web intents agent"? For
>> web content, I do see the UA as the mediating agent that gets notified
>> when web content invokes and intent, and which delivers intents to web
>> content. I'm not sure another option makes sense.
>>
>> The UA is pretty unconstrained by the spec in terms of what process
>> handles intents, though -- it is free to delegate dispatch, or adapt
>> dispatch to local OS-level processes.
>>
>>
>> On Mon, Jun 4, 2012 at 10:14 AM, Jean-Claude Dufourd
>> <jean-claude.dufourd@telecom-paristech.fr> wrote:
>>> Dear all,
>>>
>>> Web Intents, as in the current draft, is implemented entirely within the
>>> User Agent.
>>> If the UA does not know of an intent, it cannot be invoked.
>>> In a sense, the way the spec is written, the Web Intents Agent MUST be the
>>> User Agent.
>>> I think that is wrong, architecturally speaking, in the sense of
>>> "architecture of the web".
>>>
>>> I remember a discussion about a hierarchy of intent directories: if the UA
>>> does not know about an intent, it should pass the buck to a higher
>>> authority, possibly recursively.
>>> The model is that there may be an intent register at the device level, one
>>> at the home level, one at the ISP level, etc.
>>> The way the draft is written means EVERYTHING has to be done by the UA, and
>>> I do not see how the model I describe could be realized with the current
>>> draft.
>>>
>>> From a different point of view, I am sure there is a missing feature here,
>>> in the sense that the directory of intents that is meaningful to "me as a
>>> user" is NOT tied to a browser, nor to a particular computer or computing
>>> device, nor to a location (home or office), etc. So the current draft is
>>> "mistaken" in focusing everything on the UA rather than on the user.
>>>
>>> It is OK to have a default implementation of the Web Intents Agent (WIA)
>>> inside the User Agent, but it should be possible to supercede it with a more
>>> powerful implementation, or one with a more secure approach (e.g. webinos).
>>> For example, the UA-native implementation would just have one single,
>>> non-hierarchical WIA.
>>> But then, if I am an advanced user, I should be able to switch to a WIA e.g.
>>> with multiple "circles" (me, family, friends).
>>>
>>> And I should be able to share a Web Intents configuration regardless of
>>> which browser I am using at any time or on any device (specially if the
>>> device forces me to use one browser).
>>>
>>> Does this make sense to others ?
>>> Best regards
>>> JC
>>>
>>> --
>>> JC Dufourd
>>> Directeur d'Etudes/Professor
>>> Groupe Multimedia/Multimedia Group
>>> Traitement du Signal et Images/Signal and Image Processing
>>> Telecom ParisTech, 37-39 rue Dareau, 75014 Paris, France
>>> Tel: +33145817733 - Mob: +33677843843 - Fax: +33145817144
>>>
>>>
>>

Received on Tuesday, 5 June 2012 18:42:17 UTC