1. Strategies for archiving the Danish web space Bjarne Andersen Head of Digital Resources State and University Library, Aarhus bja@netarkivet.dk http://netarchive.dk

2. Agenda  New legal deposit law in Denmark  Collection strategies  NetarchiveSuite software package  Snapshot harvesting  Selective harvesting  Event harvesting  Challenges in snapshot harvesting  Snapshot harvesting usefulness  Future work

3. Legal deposit law 1  Revision of the legal deposit law in 1997 -> legal deposit included static documents on the internet  During in 1998-1999 we found out that: We were actually preserving the least interesting part  Many of the documents in that collection are also available in print  A lot of work was done between 2000-2004 2 pilot projects run by the two national libraries  Testing different software / different strategies for archiving / storing web material A governmental publication on ”preserving the Danish digital cultural heritage” (2003) A report to the ministry of culture (2004) outlining  Recommendations from the two national libraries on how to solve the ”entire” problem  Issues to be covered by a new revision of the legal deposit law

4. Legal deposit law 2  A new revision came into force on july 1st 2005 Allowing the two national libraries to automatically gather all danish websites Danish roughly defined as:  Websites on the.dk TLD  Websites minded on a Danish audience / written in Danish  Websites about Danish people (Hans Christian Andersen)  More or less any site of interest to Denmark We are by law granted access to all relevant data from the.dk TLD administrator

5. Legal deposit law 3  The law covers all public available material Material that all Danish people in principal can gain access to  Material which requires action before usage (payment, registration….)  Pay-sites should hand out username / password upon request (for free)  Other interesting parts Combined strategy (snapshot, selective and event-harvesting) Robots.txt explicitly mentioned in the regulations of the law  A lot of the very interesting websites have very restrictive robots.txt’s (we discovered around 35.000 robots.txt-files)  During 6 snap shots of more than 750.000 web sites we had fewer than 50 complaints about robots.txt

6. Legal deposit law 4  In the end led to funding of Netarchive.dk  Virtual centre in cooperation between The Royal Library, Copenhagen The State & University Library, Aarhus  Implementing a complete system  Running the archiving on a daily basis Currently with an annual budget of 450.000 euros  Involving 15 people from the two libraries 4.5 Man-years of man-power

7. The 3 collection strategies  Illustrated by coverage over time  Amount of data collected so far Snapshots: 61 TB (6 times) Selective harvests: 9.5 TB (80 web sites) Event harvests: 5.6 TB (9 events)

8. NetarchiveSuite software package  We needed a curator tool ready at July 1st 2005 Requirement number 1: Operated by librarians  With the web interface librarians can: Define harvests (all three types)  Based on quite simple settings + a number of different predefined heritrix setups Do quality control  Looking at harvest results (simple reports and statistics)  Browsing through harvested material Automated pickup of missing URIs  NetarchiveSuite was released as Open Source in July 2007 Currently used by a number of national libraries

9. Snapshot harvesting  The.dk TLD currently holds > 750.000 active domains We encountered around 42.000 Danish domains outside the.dk TLD  By extracting links from the entire.dk web space – checking country-code by IP-number (GeoIP)  By doing Google searches on Danish localities (city names..)  With 8 machines we can do One complete snapshot (including deDuplication) at 20TB in 80 days DeDuplication saves around 30% of the storage space

10. Selective harvesting  Archiving of 80 selected websites News sites ”Typical” dynamic and heavily used sites representing civic society, the commercial sector and public authorities Experimental and/or unique sites, documenting new ways of using the web (e.g. net art) Harvested much more frequent  From weekly to several times per day

11. Event harvesting  Combining the other two strategies Taking a larger number of sites (200-3000) On a more frequent basis (daily / weekly) In a shorter period of time  We have done 9 event harvests so far Elections, different national events  We have pre-defined some harvest-definitions on especially news-sites (both local and national) With one click we can start these if a sudden event should happen – to ensure collection of important sites from the very beginning

12. Challenges in snapshot harvesting  Number of domains is constantly growing 2005: 607.000 domains – 480.0o0 active 2008: 950.000 domains – 750.000 active  Domains are growing bigger and bigger Audio/Video is getting more and more popular Sites larger than 10Mb increased from 40.000 to 90.000 Sites larger than 500Mb increased from 6.000 to 12.000  Web 2.0 makes harvesting difficult Web material is inlined from other web sites – from all over the world  The border of a web site is disappearing The web is going more and more dynamic – Flash / Ajax  The amount of traps and spam grows constantly In Denmark librarians manually inspect all websites larger than 1Gb  Currently over 3000 domains  They identify aliases and potential crawler traps That task should be (semi)-automated

13. Snapshot harvesting usefulness  With snap shot harvesting a web archive ensures cultural heritage by Archiving regular ”pictures” of entire national parts of the internet Archiving as much as possible in a quite cheap way  Netarchive.dk: Storage space and 15 hours per week for librarians  Snap shots is very useful for research in many different areas Linguistics Web technologies File formats and their evolution Web design Genealogy / Ancestor search Web site history And many many more – to be defined in the future  And off cause useful for more ordinary users wanting To find content disappeared from the live web – 40-100 days lifetime  Getting more and more interesting over time  Currently access to Netarchive.dk is limited to researchers

14. Future work  Automating discovery of Danish web sites outside the.dk TLD  Automated quality assurance for large crawls  Automating filtering of web spam and traps  Improving archiving of web 2.0 Dynamic web content Streaming audio/video  Non of these problems are Danish Lets solve them together LIWA – European project working on most of these problems  Danish challenges Working for better access possibilities  On the system level: WayBack Machine / NutchWAX search  On the political level: Change of law

15. Questions ?