Re: [dita] Rationale for <alt> inside <image>

(copying this to data-lightweight as it also affects that standard)

That may explain why there is an <alt> instead of @alt, not why <alt> is inside <image> and can itself contain a <ph> which can contain an <image> etc. If <alt> would be part of <fig>, there would not have been any problem with any rendering system, just like there is no problem with <title> at the top of the <fig>, where lots of PDF renderings move the <title> to a caption below the image.

To add to the problem, LwDITA is copying the same situation, with <alt> inside <image>. With <image> essentially being an inline element and <alt> a block element, this implies that mixed content makes its way into a standard that tries to do away with it altogether. Also, there is an irrational mismatch between the semantic markup of <image> and its modern multi-media counterparts <video> and <audio>: those elements have separate child elements for setting various parameters (track, controls, source, loop, etc - none of which have translatable content), where <image> has attributes for all such configuration data plus <alt> as a child element.

A consistent semantic markup for multi-media items would use similar items for similar meanings. If <video>, <audio> and <image> are all seen as part of <fig>, they should all have a similar internal structure and markup.

My suggestion would be:

<element name=“fig”>

	<optional>

		<element name=“title”/>

	</optional>

	<optional>

		<element name=“alt”/>

	</optional>

	<choice>

		<element name=“image”>

			<attribute name=“href”/>

			<attribute name=“height”/>

			<attribute name=“width”/>

			<attribute name=“scale”/>

		</element>

		<element name=“video”>

			<attribute name=“source”/>

			<attribute name=“height”/>

			<attribute name=“width”/>

			<attribute name=“scale”/>

			<attribute name=“controls”/>

			<attribute name=“autoplay”/>

			<attribute name=“muted”/>

			<attribute name=“track”/>

			<attribute name=“poster”/>

		</element>

		<element name=“audio”>

			<attribute name=“source”/>

			<attribute name=“controls”/>

			<attribute name=“autoplay”/>

			<attribute name=“muted”/>

			<attribute name=“track”/>

			<attribute name=“poster”/>

		</element>

	</choice>

</element>

To comply with various accessibility standards, the <alt> would be made mandatory and would, obviously be required for video and audio as well.

Jang F.M. GraatSmart Information DesignAmsterdam, Netherlands

On 20 Dec 2017, at 17:09, Chris Nitchie <> wrote:

In addition, there are some HTML5-based UI frameworks that provide rich-text tooltip widgets via _javascript_ that could be fed by this element.ChrisOn 12/20/17, 10:55 AM, " on behalf of Eliot Kimber" < on behalf of > wrote:    That design comes from the original IBM design, as far as I know.    As a general principle, any content that is likely to be translated should be in element content, not attributes, which is sufficient to require <alt> rather than @alt.    Cheers,    E.    --    Eliot Kimber    http://contrext.com    On 12/20/17, 3:02 AM, "Mr. Jang Graat" < on behalf of > wrote:        I am working on a new DITA implementation for FrameMaker and stumble across a number of design issues, which make me wonder about the use cases that prompted the committee to choose this particular design.        I have not yet found a use case that would warrant the rich semantics on an alternate text for an image that is presented by <alt>. I have yet to come across an editing and rendering system that would do something useful with such rich semantics inside <alt>. AFAIK, HTML merely offers an @alt on an <img>. Would it not have been wiser to have an @alt on <image> and push the richer semantics to a sibling of <image>, as part of a <fig> ?        I really want to understand the design rationale for this and hear some convincing use cases. If such rationale and use cases are any less than convincing, I would want to propose a redesign for DITA 2.0.        Jang    ---------------------------------------------------------------------    To unsubscribe from this mail list, you must leave the OASIS TC that     generates this mail.  Follow this link to all your TCs in OASIS at:    https://www.oasis-open.org/apps/org/workgroup/portal/my_workgroups.php