Annotea futures? Annotation standards in 2009...

[Dan Brickley-2]

(I'm cc:'ing 3 lists, rather warily; if the thread gets long, please
consider trimming it to just use semantic-web@...)

Hi all

I'm involved in helping advise a new not-for-profit project that is
close in approach to the old Annotea project, looking at annotations
within pieces of Web content, and their cross-linking, threading for
discussion etc. It's now 2009, over ten years since the original Annotea
designs. The Web has changed a lot since then, but the need to annotate
it doesn't seem to have gone away.

See http://annotea.org/ 
http://www.w3.org/2001/Annotea/User/Tutorial/quicktutorial and nearby
for an overview of Annotea.

Since then Web 2.0 has happened, and now many of the original themes of
Annotea are part of the mainstream Web developer perspective. And yet
... looking at the comments to this 2007 techcrunch survey -
http://www.techcrunch.com/2007/04/10/5-ways-to-mark-up-the-web/ - I see
project after project, startup after startup, exploring this space
without any great emphasis on data exchange standards. I guess many of
them have APIs, probably a lot of them use RSS or Atom feeds. But we
certainly haven't yet to the place imagined by Annotea: an annotation
layer for the Web that allows comments, scribbles, reviews, discussions
to be freely interlinked and overlaid using open standard formats and
protocols.

So I'm mailing the relevant (and pretty quiet) lists but cc:'ing
semantic-web@... too to ask where folk thing this stuff is heading.

When is an annotation an annotation, versus a page that happens to be a
review, or happens to have as it's primary topic another page? For
annotations at the page level, it might be that mainstream RDF work
(linked data etc) has fulfilled some of the early promise of Annotea.

But for the "annotating parts of a page" scenario that lies at the heart
of many people's notion of annotations, there doesn't seem to be much
happening in terms of practical and widely adopted standards. Lots of
startups, experiments etc but they all seem to be islands. And since
annotation systems are only really interesting when you have enough
annotations to get decent coverage, this seems a pity.

Thoughts? Am I missing some developments? What would Annotea look like
if rebuilt for the Web of 2009? If it's in RDF, the query part would
just use SPARQL, and topic classification would be SKOS. What else? Is
there implementation experience from Annotea adopters and implementors
gathered somewhere? Is there consensus for example on the best bits of
information to keep if you want a robust reference to a piece of a
potentially evolving page? How well do modern Web design habits (CSS,
Ajax etc) interact with the overlay of 3rd party annotations? Is
everyone using Firefox addons, javascript bookmarklets and Web proxies
or is there some hope for a cross-browser approach on the horizon?

thanks for any suggestions, thoughts, links etc.

cheers,

Dan

[Phil Archer-3]

That's something I ought to look at too, Nikki.

Then there's POWDER which, unless something unexpected happens, will be
at Proposed Rec very soon. That's all about annotating/describing groups
of resources cf. adding annotations within specific resources but, since
the output (once processed) is RDF, it's all interoperable.

Phil.

NJ Rogers, Learning and Research Technology wrote:

> Hi Dan,
>
> You might want to look at our <http://code.google.com/p/caboto/> project
> which was a small spin-out effort from 3 projects each with a social
> software annotations aspect.
>
> We looked at use cases from different contexts and but this has not
> really been about annotating parts of a page but more about annotating
> resources (with a dedicated 'page') such as "an event", or making a
> time-based video annotation.
>
> Nikki
>
> --On 22 May 2009 16:00 +0200 Dan Brickley <danbri@...> wrote:
>
>> (I'm cc:'ing 3 lists, rather warily; if the thread gets long, please
>> consider trimming it to just use semantic-web@...)
>>
>> Hi all
>>
>> I'm involved in helping advise a new not-for-profit project that is close
>> in approach to the old Annotea project, looking at annotations within
>> pieces of Web content, and their cross-linking, threading for discussion
>> etc. It's now 2009, over ten years since the original Annotea designs.
>> The Web has changed a lot since then, but the need to annotate it doesn't
>> seem to have gone away.
>>
>> See http://annotea.org/
>> http://www.w3.org/2001/Annotea/User/Tutorial/quicktutorial and nearby for
>> an overview of Annotea.
>>
>> Since then Web 2.0 has happened, and now many of the original themes of
>> Annotea are part of the mainstream Web developer perspective. And yet ...
>> looking at the comments to this 2007 techcrunch survey -
>> http://www.techcrunch.com/2007/04/10/5-ways-to-mark-up-the-web/ - I see
>> project after project, startup after startup, exploring this space
>> without any great emphasis on data exchange standards. I guess many of
>> them have APIs, probably a lot of them use RSS or Atom feeds. But we
>> certainly haven't yet to the place imagined by Annotea: an annotation
>> layer for the Web that allows comments, scribbles, reviews, discussions
>> to be freely interlinked and overlaid using open standard formats and
>> protocols.
>>
>> So I'm mailing the relevant (and pretty quiet) lists but cc:'ing
>> semantic-web@... too to ask where folk thing this stuff is heading.
>>
>> When is an annotation an annotation, versus a page that happens to be a
>> review, or happens to have as it's primary topic another page? For
>> annotations at the page level, it might be that mainstream RDF work
>> (linked data etc) has fulfilled some of the early promise of Annotea.
>>
>> But for the "annotating parts of a page" scenario that lies at the heart
>> of many people's notion of annotations, there doesn't seem to be much
>> happening in terms of practical and widely adopted standards. Lots of
>> startups, experiments etc but they all seem to be islands. And since
>> annotation systems are only really interesting when you have enough
>> annotations to get decent coverage, this seems a pity.
>>
>> Thoughts? Am I missing some developments? What would Annotea look like if
>> rebuilt for the Web of 2009? If it's in RDF, the query part would just
>> use SPARQL, and topic classification would be SKOS. What else? Is there
>> implementation experience from Annotea adopters and implementors gathered
>> somewhere? Is there consensus for example on the best bits of information
>> to keep if you want a robust reference to a piece of a potentially
>> evolving page? How well do modern Web design habits (CSS, Ajax etc)
>> interact with the overlay of 3rd party annotations? Is everyone using
>> Firefox addons, javascript bookmarklets and Web proxies or is there some
>> hope for a cross-browser approach on the horizon?
>>
>> thanks for any suggestions, thoughts, links etc.
>>
>> cheers,
>>
>> Dan
>>
>>
>>
>>
>
>
>
> ----------------------
> NJ Rogers, Technical Researcher
> (Senior Technical Developer and Coordinator of Web Futures)
> Institute for Learning and Research Technology (ILRT)
> Email:nikki.rogers@...
> Tel: [phone] (Direct)
> Tel: [phone] (Office)
>
>

--

Phil Archer
http://philarcher.org/www@20/

i-sieve technologies                |      W3C Mobile Web Initiative
Making Sense of the Buzz            |      www.w3.org/Mobile

[Dan Brickley-2]

Thanks Nikki, Phil. Good points!

This also reminds me that Media annotations is moving along nicely over
in a separate W3C group. See http://www.w3.org/2008/WebVideo/Annotations/

Also I've just signed up for a fresh Annotea account, as I wanted to try
Annozilla (http://annozilla.mozdev.org/). It seems all the w3.org signup
machine is still working, which was a pleasant suprise. It sets a user
up with username and password for posting annotations. So - thinking
again about how something like this would be built with 2009-era specs,
I suspect OAuth might be used here. This would allow clients to be
delegated an access token for reading/posting etc annotations. At the
moment Annotea assumes each user has an account and the password for
that account is directly shared with the apps that can post to it.
Perhaps there could be benefit in having the apps (whether desktop,
in-browser or website-based) use oauth tokens instead? Or perhaps I'm
just being trendy and trying to use too many shiny new things? I do
think that AtomPub+Oauth is worth investigating, despite premature
reports (http://norman.walsh.name/2009/05/07/timing) of it's death...

cheers,

Dan

[Matthew Wilson-4]

Dan Brickley wrote:
> (I'm cc:'ing 3 lists, rather warily; if the thread gets long, please
> consider trimming it to just use semantic-web@...)
>
> Hi all
>

> Thoughts? Am I missing some developments? What would Annotea look like
> if rebuilt for the Web of 2009? If it's in RDF, the query part would
> just use SPARQL, and topic classification would be SKOS.

IMO the use of RDF seems to add a significant "complexity tax" on
implementations.

 > What else? Is
> there implementation experience from Annotea adopters and implementors
> gathered somewhere? Is there consensus for example on the best bits of
> information to keep if you want a robust reference to a piece of a
> potentially evolving page? How well do modern Web design habits (CSS,
> Ajax etc) interact with the overlay of 3rd party annotations? Is
> everyone using Firefox addons, javascript bookmarklets and Web proxies
> or is there some hope for a cross-browser approach on the horizon?

As an implementer, it seems to me that XPointer is not a great solution
for determining a selection of a web page. Theoretically it's only
specified for use with XML and not with HTML. Annotea glosses over this
problem, but there are real compatibility questions which I haven't seen
answered definitively (for example, if you have an 'implied' element not
present in the markup like "tbody", is it present in a constructed
XPointer)?

 > How well do modern Web design habits (CSS,
 > Ajax etc) interact with the overlay of 3rd party annotations?

Arguably Annozilla doesn't even work well with less modern Web design
(the hacks it performs in order to display icons in the document are
pretty horrible), but it doesn't seem to have caused many problems in
practice - or at least I haven't had many reported to me. My guess is
that the use of Annozilla is pretty limited and that it isn't getting
any widespread use on any pages with significant Ajax usage. It's
obviously trivial to create an Ajaxy page which would expose the
limitations of the schema, and you would imagine that real-life usage
would have the same difficulties.

Matthew

[Dan Brickley-2]

On 22/5/09 18:24, Matthew Wilson wrote:

> Dan Brickley wrote:
>> (I'm cc:'ing 3 lists, rather warily; if the thread gets long, please
>> consider trimming it to just use semantic-web@...)
>>
>> Thoughts? Am I missing some developments? What would Annotea look like
>> if rebuilt for the Web of 2009? If it's in RDF, the query part would
>> just use SPARQL, and topic classification would be SKOS.
>
> IMO the use of RDF seems to add a significant "complexity tax" on
> implementations.

Worth noting, and going into the practical details. Were you working
solely with the Mozilla RDF APIs? XUL Templates etc? Or other more
modern RDF libraries?

>  > What else? Is
>> there implementation experience from Annotea adopters and implementors
>> gathered somewhere? Is there consensus for example on the best bits of
>> information to keep if you want a robust reference to a piece of a
>> potentially evolving page? How well do modern Web design habits (CSS,
>> Ajax etc) interact with the overlay of 3rd party annotations? Is
>> everyone using Firefox addons, javascript bookmarklets and Web proxies
>> or is there some hope for a cross-browser approach on the horizon?
>
> As an implementer, it seems to me that XPointer is not a great solution
> for determining a selection of a web page. Theoretically it's only
> specified for use with XML and not with HTML. Annotea glosses over this
> problem, but there are real compatibility questions which I haven't seen
> answered definitively (for example, if you have an 'implied' element not
> present in the markup like "tbody", is it present in a constructed
> XPointer)?

Yup. This might be worth taking up with the HTML5 and WHATWG folks,
since they're trying to write a spec that has a recovery model for ugly
messy markup.

>  > How well do modern Web design habits (CSS,
>  > Ajax etc) interact with the overlay of 3rd party annotations?
>
> Arguably Annozilla doesn't even work well with less modern Web design
> (the hacks it performs in order to display icons in the document are
> pretty horrible), but it doesn't seem to have caused many problems in
> practice - or at least I haven't had many reported to me.

If there aren't many problems, in what sense does it not perform well?
(internal Engineering uglyness, or problems that will affect users?)

> My guess is that the use of Annozilla is pretty limited and that it isn't getting
> any widespread use on any pages with significant Ajax usage. It's
> obviously trivial to create an Ajaxy page which would expose the
> limitations of the schema, and you would imagine that real-life usage
> would have the same difficulties.

Yep. Perhaps the pages that are problematic that way might also be
problematic in terms of assessibility, and Mobile Web -readyness too?
Which would at least give authors other motivations to fix their markup,
apart from annotate-ability.

cheers,

Dan

[Matthew Wilson-4]

Dan Brickley wrote:

> On 22/5/09 18:24, Matthew Wilson wrote:
>> Dan Brickley wrote:
>>> (I'm cc:'ing 3 lists, rather warily; if the thread gets long, please
>>> consider trimming it to just use semantic-web@...)
>>>
>>> Thoughts? Am I missing some developments? What would Annotea look like
>>> if rebuilt for the Web of 2009? If it's in RDF, the query part would
>>> just use SPARQL, and topic classification would be SKOS.
>>
>> IMO the use of RDF seems to add a significant "complexity tax" on
>> implementations.
>
> Worth noting, and going into the practical details. Were you working
> solely with the Mozilla RDF APIs? XUL Templates etc? Or other more
> modern RDF libraries?

Mozilla APIs. But I also have bad memories of debugging responses from
the server (and trying to read the RDF Schema spec). As a non-expert, I
see RDF in the same category as XML Schema - trying to do much and
failing to "make the easy things easy", at least in the context of
annotations.

>>  > What else? Is
>>> there implementation experience from Annotea adopters and implementors
>>> gathered somewhere? Is there consensus for example on the best bits of
>>> information to keep if you want a robust reference to a piece of a
>>> potentially evolving page? How well do modern Web design habits (CSS,
>>> Ajax etc) interact with the overlay of 3rd party annotations? Is
>>> everyone using Firefox addons, javascript bookmarklets and Web proxies
>>> or is there some hope for a cross-browser approach on the horizon?
>>
>> As an implementer, it seems to me that XPointer is not a great solution
>> for determining a selection of a web page. Theoretically it's only
>> specified for use with XML and not with HTML. Annotea glosses over this
>> problem, but there are real compatibility questions which I haven't seen
>> answered definitively (for example, if you have an 'implied' element not
>> present in the markup like "tbody", is it present in a constructed
>> XPointer)?
>
> Yup. This might be worth taking up with the HTML5 and WHATWG folks,
> since they're trying to write a spec that has a recovery model for ugly
> messy markup.
>
>>  > How well do modern Web design habits (CSS,
>>  > Ajax etc) interact with the overlay of 3rd party annotations?
>>
>> Arguably Annozilla doesn't even work well with less modern Web design
>> (the hacks it performs in order to display icons in the document are
>> pretty horrible), but it doesn't seem to have caused many problems in
>> practice - or at least I haven't had many reported to me.
>
> If there aren't many problems, in what sense does it not perform well?
> (internal Engineering uglyness, or problems that will affect users?)

Annozilla makes internal changes to the DOM so that it can add icons and
highlighting to the document, creating spans and images. It's easy to
imagine stylesheets or scripts breaking as a result. (I know Mozilla has
XBL but it doesn't quite seem to fit my needs here, at least last time I
checked.)

>> My guess is that the use of Annozilla is pretty limited and that it
>> isn't getting
>> any widespread use on any pages with significant Ajax usage. It's
>> obviously trivial to create an Ajaxy page which would expose the
>> limitations of the schema, and you would imagine that real-life usage
>> would have the same difficulties.
>
> Yep. Perhaps the pages that are problematic that way might also be
> problematic in terms of assessibility, and Mobile Web -readyness too?
> Which would at least give authors other motivations to fix their markup,
> apart from annotate-ability.

This seems a bit optimistic.

Matthew