Archiving the Web during unforeseen events (terrorist attacks, war...)

Web content and social networks are highly ephemeral and claim for fast reactions in case of disruptive events, in order to preserve them. As demonstrated with the Yugoslavian wars, national domains can disappear (Ben-David, 2016). Web archiving and the practice of collecting and preserving born digital content have been widely adopted by national libraries in Europe (in France, Great Britain, Luxembourg, Denmark and many other countries) since the 2000s (Brügger, 2018). These collections can take many forms, including annual crawls of a country's entire webosphere, as well as special collections created for specific foreseen events such as elections, or unforeseen events like terrorist attacks, natural disasters or sanitary crisis (i.e., the terrorist attacks in Paris in 2015-16, the COVID-19 crisis, etc.). Additionally, other stakeholders are involved in preserving digital traces, such as Documenting the Now for the Black Lives Matter movement and the Sucho initiative, launched at the very beginning of the Ukrainian war and that aims to save the online cultural heritage of Ukraine. This presentation will firs examine the processes and challenges involved in creating "live archives" (Rhodes, 2013; Rollasson & Reed, 2015) during times of tensions, disruption, attacks, and war. It will explore the unique features of these collections, relying on various case studies, such as the Paris Terrorist Attacks collection at the BnF (Bibliothèque nationale de France) and Ina (French Audio-visual institute) (Schafer et al. 2017), the approach of Documenting the Now (http://www.docnow.io) and Sucho (Saving Ukrainian Cultural Heritage Online, https://www.sucho.org), and international collaborations during the COVID-19 crisis (thanks to the International Internet Preservation Consortium, https://archive- it.org/collections/13529). The presentation will then examine the implications of these collections for researchers (and notably for historians), including issues of representativeness, scope, temporalities, noise and gaps, context, and metadata.