Glad the idea appeals to you. There is much work yet to be done in this
area. We're having some difficulty working out an ontology for
representing metadata. To date, OWL is presenting the most likely path
(OWL-DL in particular). For now, the metadata would represent specific
characteristics of the delivery context, such as device info, user
preferences etc. The idea of "role" to capture the intent of markup
will probably follow.
One could envisage, for example, an embedded fragment of JavaScript
that comes with metadata to say:
1. The script represents a "link highlighting" mechanism (i.e. this is its
role).
2. The script is associated with another section whose role is "main menu".
3. The script is optional (ie omission will not cause loss of functionality)
An adaptation process will then be able to understand the purpose of
the script without having to interpret the script (which is an
Artificial Intelligence activity), and can substitute alternative
implementations of the role (e.g. vocal cues for screen-readers, or
other implementations for non-JS browsers).
I can see why you would think JSF, and certainly in the case of forms
and other interaction components this makes sense. But the bigger
picture is to represent the author's intent at varying levels of
granularity. For example, to mark fragments of a document that
represent "essential information", "key message", "optional information
(side-bars, footnotes etc)", "search (site|nearby|related-to-session|...)",
"error", "explanation of error", "most likely next step", "illustration",
"decorative resource" etc etc etc
I'm just making up these roles, because so far we have only identified
that "role" is important but don't yet have a process for deciding such
roles. Nor is there a representation. There is a strong feeling that
RDF/OWL will be playing a big part.
Obviously what I'm thinking about is the annotation of methods/classes
in Beehive. If the UI could conveny some of its inherent role
information through metadata, the possibilities for the Semantic Web
and content adaptation could be huge. Propagating the metadata up to
the UI would be an interesting avenue to explore, but it wouldn't be
unusual. How many times have you viewed the source of a news web page
and seen embedded comments that say things like:
<!-- Main News Story --> ?
Wouldn't it be great if this accidental leakage of metadata
wasn't just accidental, but was intentional, and instead of human
comments it was machine-readable?
Yes, you could put the annotation into the JSP directly. But imagine
the JSP could receive this information from the underlying application.
Not only would the world of PDAs and mobile/cell-phones be
beneficiaries of the knock-on effects, but the millions of end-users
who have special needs (the blind, those with mobiliy challenges etc.)
would finally have access to a Web that was adaptable in a sensible
manner.
So, I look at Beehive and I see that embedded within the code is
potentially the machine-readable metadata that could enable adaptation
processes make sense of what it "sees" and thereby adapt it sensibly to
the needs of its users and user-agents. And the adaptation could take
place anywhere where the metadata is available, from server to client.
Because of the potential, I will make a point of keeping the community
informed of developments. But please be patient because experience
shows that this is likely to be a slow process.
---Rotan.
-----Original Message-----
From: Richard Feit [mailto:[EMAIL PROTECTED]
Sent: Wed 20/10/2004 18:14
To: Beehive Developers
Cc:
Subject: Re: Role - an adaptation metadata/annotation
This is an interesting idea. Currently, there's never a control that's
wholly responsible for rendering a piece of UI (at least that's not the
typical pattern). Often there's a control involved upstream in
providing data to back the UI, but in the end the NetUI tags in a JSP
are the final word. Seems like in this case, RDF-type information would
still be embedded in the JSP using the same mechanism that evolves for
web pages in general. I can actually see where this might fit better
with technology like JSF, where there are actual UI component classes
that could be annotated in the way you're suggesting.
That said, I think it would be great for us to look at standardized
semantic metadata across the board (web services, controls, UI). If
you've got ideas for how this metadata would propagate from the control
layer to the UI, I'd definitely be interested in exploring that.
Rich
Rotan Hanrahan wrote:
>This is just a quick update on some W3C work I've been doing recently.
>(See: http://www.w3.org/2004/06/DI-MCA-WS/)
>
>We were looking for metadata concepts that would be useful in the
>adaptation of web content for multiple devices. The diverse group of
>participants agreed a few points; one in particular may be of interest
>to Beehive.
>
>It was agreed that a web page could include metadata marking subsets of
>the document according to the "role" that is played by that subset. A
>number of technologies for implementing such a mechanism have been
>considered (OWL, RDF and other SemWeb concepts), but obviously it could
>be "a while" before this is translated into a tangible proposal.
>
>However, it occurred to me that if a Beehive control, responsible for
>the generation of a subset of the web page, could be annotated with the
>same ontology applied to "role" then that control could also express
>the metadata (e.g. as embedded RDF) to facilitate subsequent adaptation
>of that content.
>
>Simple example: a control that generates the main menu of the page
>would not only generate the markup but would also generate embedded RDF
>to indicate that the markup was in fact the "main menu" and not some
>random/unrelated collection of links. This information might later be
>used by content adaptation (e.g. client-side) to reposition the main
>menu for a different device/browser.
>
>Getting such an idea into the mainstream could take a long time, but
>the existence of a respected software platform capable of supporting
>such a concept might help oil the wheels of standardization bodies...
>
>I will be updating the W3C workshop pages over the next few weeks.
>
>---Rotan
>
>
