Not all digital formats are suited or indeed designed for archiving or preservation. Any preservation policy should therefore recognize the requirements of the collection content and decide upon a file format which best preserves those qualities. Pairing content with a suitable choice of preservation format or access format; identifying what is important in the content.
Below we suggest some factors to consider in selecting your preferred file formats:
Open source formats, such as JPEG2000, are very popular due to their non-proprietary nature and the sense of ownership that stakeholders can attain with their use. However, the choice of open source versus proprietary formats is not that simple and needs to be looked at closely. Proprietary formats, such as TIFF, are seen as being very robust; however, these formats will ultimately be susceptible to upgrade issues and obsolescence if the owner goes out of business or develops a new alternative. Similarly, open source formats can be seen as technologically neutral, being non-reliant on business models for their development however they can also been seen as vulnerable to the susceptibilities of the communities that support them.
Although such non-proprietary formats can be selected for many resource types this is not universally the case. For many new areas and applications, e.g. Geographical Information Systems or Virtual Reality only proprietary formats are available. In such cases a crucial factor will be the export formats supported to allow data to be moved out of (or into) these proprietary environments.
The availability of documentation - for example, published specifications - is an important factor in selecting a file format. Documentation may exist in the form of vendor’s specifications, an international standard, or may be created and maintained within the context of a user community. Look for a standard which is well-documented and widely implemented. Make sure the standard is listed in the PRONOM file format registry.
A file format which is relied upon by a large user group creates many more options for its users. It is worth bearing in mind levels of use and support for formats in the wider world, but also finding out what organizations similar to you are doing and sharing best practice in the selection of formats. Wide adoption of a format can give you more confidence in its preservation.
Lossy formats are those where data is compressed, or thrown away, as part of the encoding. The MP3 format is widely used for commercial distribution of music files over the web, because the lossy encoding process results in smaller file sizes.
TIFF is one example of an image format that is capable of supporting lossless data. It could hold a high-resolution image. JPEG is an example of a lossy image file format. Its versatility, and small file size, makes it a suitable choice for creating an access copy of an image of smaller size for transmission over a network. It would not be appropriate to store the JPEG image as both the access and archival format because of the irretrievable data loss this would involve.
One rule of thumb could be to choose lossless formats for the creation and storage of "archival masters"; lossy formats should only be used for delivery / access purposes, and not considered to be archival. A rule like this is particularly suitable for a digitization project, particularly still images.
Some file formats have support for metadata.This means that some metadata can be inscribed directly into an instance of a file (for example, JPEG2000 supports some rights metadata fields). This can be a consideration, depending on your approach to metadata management.
This is a complex area. One view regards significant properties as the "essence" of file content; a strategy that gets to the heart of "what to preserve". What does the user community expect from the rendition? What aspects of the original are you trying to preserve? This strategy could mean you don’t have to commit to preserving all aspects of a file format, only those that have the most meaning and value to the user.
Significant properties may also refer to a very specific range of technical metadata that is required to be present in order for a file to be rendered (e.g. image width). Some migration tools may strip out this metadata, or it may become lost through other curation actions in the repository. The preservation strategy needs to prevent this loss happening. It thus becomes important to identify, extract, store and preserve significant properties at early stage of the preservation process.