The sizes of the viewport area and the content area are defined in terms of lengths (width and depth).

Lengths must be expressed as a decimal value followed immediately by an optional unit of measure or a percentage. Six and one eighth inches, for example, must be expressed as “6.125in”. It is an error to put a space or other punctuation between the decimal value and the unit of measure.

Examples of common units of measure include:

If no unit of measure is provided, px is assumed. Note that pixels have no universally accepted absolute size and ems are relative units of measure. Implementations may define pixel sizes differently and stylesheets may or may not be able to determine the current font size in order to correctly calculate the absolute size of an em. It is best to avoid these units of measure.

Percentages are expressed as a decimal value followed immediately by a % sign.

The viewport area is specified by the width and depth attributes.

If neither width nor depth is specified, an implementation is free to choose defaults. These defaults may be influenced by context. For example, when rendering an inline graphic, the viewport area often defaults to the size of the content area. For block graphics, the width often defaults to the column width while the depth defaults to the depth of the content area.

If only one of width or depth is specified, an implementation is free to choose a default for the other dimension.

Viewport area dimensions expressed as a percentage are a percentage of the available area. For example, a width of 50% when an implementation is rendering in a column 6in wide is equivalent to specifying a width of 3in.

Percentages must be used with care. Some media are unbounded in one or more directions (e.g., web pages are generally unbounded in depth). Specifying a percentage of an unbounded dimension is undefined. Implementations may choose arbitrary defaults or may generate errors.

The content area is specified by the contentwidth and contentdepth attributes.

If neither content width nor content depth is specified, an implementation is expected to render the image at its intrinsic size (unless scaling or scaling to fit is requested). If only one of content width or content depth is specified, an implementation is expected to choose a default for the other dimension such that the image is scaled proportionally. For example, if an image has an intrinsic size of one square inch and the content width is specified as 2in, the content depth must default to 2in.

Content area dimensions expressed as a percentage are a percentage of the intrinsic size of the image.

Percentages must be used with care. Some implementations may be unable to determine the intrinsic size of an image and will therefore be forced to make compromises. Implementations may choose arbitrary values or may generate errors if the intrinsic size cannot be obtained.

Scaling can be specified in two ways: with the scale attribute or with the scalefit attribute.

If scale is specified, it must be a positive integer. It is always interpreted to be a percentage value, where 100 represents 100%.

The legal values of scalefit are 0 (false) or 1 (true). If scaling to fit is requested, the content area is scaled until either the content width is the same as the viewport width (and the content depth is less than or equal to the viewport depth) or the content depth is the same as the viewport depth (and the content width is less than or equal to the viewport width), whichever comes first. In other words, scaling to fit never causes anamorphic scaling; it simply scales the image as large as possible without overflowing the bounds of the viewport area.

Specification of content area, scaling, and scaling to fit are mutually exclusive. If a content area (contentwidth, contentdepth, or both) is specified, both scaling and scaling to fit are ignored. If the content area is not specified and both scaling and scaling to fit are specified, scalefit is ignored.

In order to achieve a level of backward compatibility with previous versions of DocBook (which did not have attributes for specifying a content area) while maintaining coherent semantics, the default value of scalefit depends on other attributes.

If a viewport area is specified (and neither a content area nor scaling is specified) and scalefit is explicitly 0, the viewport area specification must be ignored, and the image must not be resized.

Two alignment attributes are provided: align and valign.

If specified, align indicates how the content area should be aligned horizontally within the viewport area. If not specified, implementations are free to choose any default value.

If specified, valign indicates how the content area should be aligned vertically within the viewport area. If not specified, implementations are free to choose any default value.

If nothing is specified about the size of an image, it is rendered in a content area that is the same as its intrinsic size in a viewport area that is implementation defined:

<imagedata fileref="image.png"/>

If a viewport area is specified, the image is rendered in a content area that is the same as its intrinsic size in the specified viewport area:

<imagedata fileref="image.png" width="6in" depth="5.5in" scalefit="0"/>

If a content area is specified, the image is scaled (possibly anamorphically) to that size and rendered in a viewport area that is implementation defined:

<imagedata fileref="image.png" contentwidth="4in" contentdepth="3in"/>

If a scaling factor is specified, the intrinsic size is scaled uniformly by that amount to obtain the content area which is rendered in a viewport area that is implementation defined:

<imagedata fileref="image.png" scale="300"/>

If a viewport area is specified and scaling to fit is requested, the intrinsic size is scaled (uniformly) as large as possible without extending beyond the bounds of the viewport area which is rendered as specified:

<imagedata fileref="image.png" width="6in" depth="5.5in"/>
<!-- note that scalefit="1" is the default in this case -->

If the viewport area and content area are specified, the image is scaled (possibly anamorphically) to the content area size and rendered in the specified viewport area:

<imagedata fileref="image.png" width="6in" depth="5.5in"
           contentwidth="4in" contentdepth="3in"/>

If the viewport area and a scaling factor are specified, the intrinsic size is scaled uniformly by the scaling factor amount to obtain the content area which is rendered in the specified viewport area:

<imagedata fileref="image.png" width="6in" depth="5.5in" scale="300"/>