W3C home > Mailing lists > Public > public-html@w3.org > April 2011

Re: longdesc - beside the box

From: Benjamin Hawkes-Lewis <bhawkeslewis@googlemail.com>
Date: Tue, 26 Apr 2011 02:25:35 +0100
Message-ID: <BANLkTimJeHD_wS1CaZbMRx9nLiQPnwB45A@mail.gmail.com>
To: John Foliot <jfoliot@stanford.edu>
Cc: Leif Halvard Silli <xn--mlform-iua@xn--mlform-iua.no>, HTMLWG WG <public-html@w3.org>
On Mon, Apr 25, 2011 at 8:31 PM, John Foliot <jfoliot@stanford.edu> wrote:
> *IT's NOT* hidden metadata, it's discoverable metadata

"Hidden" and "discoverable" are not exclusive categories.

> just like microformats are discoverable metadata: I've yet to encounter a
> browser or web-page that renders microformat data on screen.

On the contrary, rendering microformat data to the screen is the expected norm.

http://microformats.org/wiki/principles

For example, consider rel="tag":

http://microformats.org/wiki/rel-tag

An implementation exemplifying visible data would be a blogpost followed by a
visible link to a related tag:

  <a href="http://technorati.com/tag/tech" rel="tag">tech</a>

The meaning of the microformat (i.e. this post is about tech) is communicated to
humans in the default visual presentation, and available to aggregators as
machine-readable markup.

An example of markup typifying the hidden metadata anti-pattern would be:

  <meta name="keywords" content="tech">

Here the metadata is hidden from humans in the default visual presentation,
although it is still available to aggregators.

To avoid hidden metadata, you must express the content and relationships, such
as the tag for a post or the long description for an image, through a "default
visual encumbrance", to use Laura's phrase. For example, this is visible data:

  <img src="chart.jpg" alt="Sales rose 24%" aria-describedby="desc">
  <details>
    <summary>Long description</summary>
    <p id="desc">[long description of chart goes here]</p>
  </details>

And this is hidden metadata:

    <img src="chart.jpg" alt="Sales rose 24%" longdesc="somewhere">

> Yet user-agents can none-the-less discover this data, and a number of browser
> plug-ins (such as the Operator and Tails Export extensions for firefox) allow
> individual users the ability to extract this discoverable metadata into a
> usable form they can (visually) process.

In the case of publisher implementations living up to the visible data
principle, such tools merely extract information users can already see and
represent it in a new form, rather than making hidden information visible.

The general idea of microformats is not to hide data so that it can be made
visible through tools, but to markup visible information so that it can be
aggregated or otherwise repurposed (e.g. dragging contact information from a
web page to an address book application).

I do not think anyone who understands the visible data principle will change
their minds about @longdesc in response to attempts to rebrand "hidden
metadata" as "discoverable metadata". Rather the benefits of "longdesc" must be
shown to outweigh the costs of the antipattern - namely, that authors are less
likely to spot errors in hidden metadata than visible data.

http://tantek.com/log/2005/06.html#d03t2359

This depends on "longdesc" substantially increasing the number of descriptions
authors are willing to provide, and on better tools (validators, browsers,
etc.) substantially reducing the error rate documented in:

http://blog.whatwg.org/the-longdesc-lottery

--
Benjamin Hawkes-Lewis
Received on Tuesday, 26 April 2011 01:26:03 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Wednesday, 9 May 2012 00:17:28 GMT