A dark archive is an archive that is not meant to be publicly accessible. As you said it is used to store information that must survive in case of disaster.
But what is a disaster for a library? Yesterday it was a fire, today is a publisher going bankrupt and dismissing its online (paywalled) archive. A dark archive pays the fee to access various online resources and store them indefinitely, but without letting anyone access that information (well, except the archive admins). The legality of this? Dubious. The usefulness of this? High.
As Portico.com puts it, «We chose to create a “dark” archive to focus our efforts on securing and preserving large volumes of content important to libraries and their users; however, it is not exclusively dark. Participating libraries experience the archive as a “light” or accessible archive in two ways: auditing the archive to ensure we are prepared to support eventual use and accessing of content that has been made available as the result of a “trigger event” or post-cancellation access claim. Unlike many ongoing preservation initiatives, Portico participants and their users experience direct customer support, should they ever need it.»
But the devil is in the details. Who pays the fee for these dark archives? Do we trust them? Who is auditing what is happening behind the doors of the dark archives? Luckily there are FLOSS systems like LOCKSS and CLOCKSS that allow you to create your own dark archive.
Few links:
* http://news.jstor.org/news/2004.12/repositories.html
* http://hul.harvard.edu/publications/ln1356/03.html
* http://www.clockss.org/clockss/FAQ
* http://www.clockss.org/clockss/How_CLOCKSS_Works
* http://www.portico.org/digital-preservation/the-archive-cont...