Large-scale collections under the magnifying glass : Format identification for web archives

International audienceInstitutions that perform web crawls in order to gather heritage collections have millions - or even billions - of files encoded in thousands of different formats about which they barely know anything. Many of these heritage institutions are members of the International Internet Preservation Consortium, whose Preservation Working Group decided to address the issues related to format identification in web archive. Its first goal is to design an overview of the formats to be found in different types of collections (large-, small-scale...) over time. It shows that the web seems to be becoming a more standardized space. A small number of formats - frequently open - cover from 90 to 95% of web archive collections, and we can reasonably hope to find preservation strategies for them. However, this survey is mainly built on a source - the MIME type of the file sent in the server response - that gives good statistical trends but is not fully reliable for every file. This is the reason why it appears necessary to study how to use, for web archives, identification tools developed for other kinds of digital assets