Not Just Decoration: quality graphics for the Web

Embedded metadata

The problem

The topic of metadata has received some attention in connection with textual data. Graphics too can have important metadata, such as Copyright details, descriptions of the content of the image, legal disclaimers, or technical descriptions of the method of image synthesis employed. More application-specific metadata could include:

and so on. This information needs to be associated with the graphical information in some way. As images become used as more than mere adornment, the role of image metadata will grow. Quality graphics have associated metadata.

To embed or not to embed

The traditional method of transmitting such information has been via a separate text file. There is however the risk that the two files will become separated. Metadata can also be supplied as an HTML file, with the image linked to or supplied inline, to make a stronger connection between the image and it's metadata.

These methods have consistency problems - particularly in a cacheing proxy environment - when either the image or the metadata needs to be altered.

Rather than embedding the image in the metadata, several image formats allow text of various kinds to be contained within an image. The advantage of this is that the metadata cannot be inadvertantly separated, and hence for example the copyright details of a graphic are unambiguous. Keeping the information in a single file also permits the use of message digests (as proposed in the draft HTML 3.0 specification [19]) on link anchors, to ensure that the currently linked-to graphic is exactly what the document author intended to link to with that URL.

The disadvantage of embedded metadata is that the Web user agent must extract the information and display it. The means of doing so are necessarily different for each Internet Media type. It is a trade off between inconvenience to the implementor and the convenience to the user of a single consistent user interface. This has to date been one of the Web's strengths, and a browser which presented embedded metadata by generating an HTML page would be highly suited to applications where image quality was considered important.

A third, compromise possibility would be to link to a multipart Internet Media type containing image data and an accompanying metadata file. There is as yet little experience, however, with the presentational aspects of multipart objects.

Notes to implementors


There are no known interactions between providing this facility and any other aspects of graphical excellence.


Providing the ability to directly view embedded metadata in graphics files for a small number of suitable Internet Media types associated with high quality graphics would be an interesting experiment. PNG is a relatively simple format which provides for text chunks with associated keywords, and the coded character set for these chunks is ISO Latin-1. There is a freely available C library for reading and writing PNG files, including extraction of text chunks. It would thus be a suitable candidate for trial implementation of this facility.

Next: Conclusions and References