The document discusses tagfs, a tagging semantics for hierarchical file systems, and highlights its advantages over traditional file systems, such as improved organization and flexibility. It introduces the concept of semfs, which implements tagfs using a semantic approach, allowing for better query handling and resource management. The paper also addresses challenges, such as file renaming and identity, while providing insights into the architecture and potential applications of these systems.

1. TagFS — Tag Semantics for Hierarchical File Systems Bonus Track: Introducing SemFS 2006 2006 Web3.0 2006 Stephan Bloehdorn Institute AIFB, University of Karlsruhe, Germany Olaf Görlitz ISWeb, University of Koblenz-Landau, Germany Simon Schenk ISWeb, University of Koblenz-Landau, Germany Max Völkel Forschungszentrum Informatik, Karlsruhe, Germany talk Max Völkel Forschungszentrum Informatik, Karlsruhe, Germany

2. Motivation „ Every user [...] indicated that their attempts to establish elaborate filing schemas for archived information failed because they proved to require more time and effort than the information was worth.“ Barreau and Nardi, 1995 Contributions Mapping file system semantics to tagging semantics Architecture for a semantic file system

3. Hierarchical file systems have some problems Single locaction property E.g. where to put a song produced by two artists? Each file may be only in one folder There are links/shortcuts, but where to put the „primary“ file? Browsing to maximum specifity E.g. 5 clicks for /My Music/Fatboy Slim/2006/danceable/favourite even if there are only 5 Fatboy Slim songs altogether The more you organise, the more you have to browse Missing orthogonality E.g. /2003/Fatboy Slim/favourite or /favourite/Fatboy Slim/2003 ? Many dimensions, only one access path No query refinement FS lists only directories explicitly placed there, no help

4. Tagging Simple idea: Instead of putting resources in folders (nested containers), put tags (labels) on resources Tagging user interfaces User sees a resource, can type in tags (simple, single keywords, separated by space or comma) User can click on a tag, UI lists all resources with that tag assigned Conjunctive queries: e.g. fatboyslim+favourite Examples: del.icou.us, flickr, 100 more Who has used a tagging system? new Web2.0

5. Example: del.icio.us Tagging Browsing Queries

6. Comparison File system: partition of the adress space Tagging: overlapping sets a b a+b+c a+c b+c a+b c a b c /a /a/b /a/c /a /a/b /a/b/c /a/c /a/c/b /c /c/b /c/b/a /c/a /c/a/b /b /b/a /b/a/c /b/c /b/c/a

7. Mapping file system semantics to tagging semantics Query and Browse – the easy parts Query Use path as query, e.g. /a/b/c = query for a + b + c Browsing / a Contained files: all resources tagged with a (  if not to many) Contained folders: All tags b , for which the conjunctive query a+b is not empty Any tag is a good starting point! Note: Virtual directory views are computed at runtime

8. Mapping file system semantics to tagging semantics Tagging – the hard part Changing existing tagging Copy file from a to b = tag file with b also Move file from a to b = remove tag a , add tag b Delete file from a = remove tag a Rename a to b = for all files tagged with a : remove a , add b Add files to TagFS Add file to folder a = add file to TagFS; tag file with a  File identity determined by hash or filename  Allows updating a file, if content is changed externally Delete files from TagFS  Move file to deleteMe = delete file Create tag a = create folder a  This folder would not be shown, because it‘s empty

9. Tagging in the file system Many locaction property Each file may carry as many tags as desired Browsing until result is small enough Each folder contains all files tagged with the folder name Orthogonality of information dimensions Path interpreted as a conjunctive query, e.g. Lisa Ekdahl/2006 is the same as 2006/Lisa Ekdahl Not all reviewers agreed on this  Query refinement Each folder lists useful query refinements as sub-folers Providing all standard filesystem operations  compatibility with existing applications

10. Introducing SemFS Implementing TagFS with SemFS Web2.0 Web3.0 Semantic Desktop RDF keywords Bonus Track

11. What is a file system?

12. What is a file system? Organising files … An address given as path expression := Letter “ : “ ( “ \ “ name)* File System dir! response files folders name metadata metadata

13. What is a file system? … managing binary data File System metadata data 1011 write 1011 read 1011

14. What is a file system? … managing binary data File System rename metadata data

15. What is a file system? … managing binary data File System Add file or delete file metadata data

16. What is a file system? Organising files and managing binary data View( path )  Metadata-Table Contained files, Contained directories, Metadata: getName, getDate, … Binary content irrelevant Update Metadata: setName, setDate, add file, delete file, move file Binary+Metadata: read file, write file, trunc file

17. What is a virtual file system? Organising files and managing binary data op Virtual File System Looks and behaves like a file system… … but is no file system. It‘s implemen-ted differently. metadata data X Flickr CMS File System metadata data

18. What is a semantic file system? Organising files and managing binary data. op Semantic File System Flexible implementation. Unified metadata  unified search metadata data X Flickr CMS File System

19. Architecture of SemFS: Filters Input: Frontend gets path SemFS maintains a list of filters , Formal: Filter(graph, path)  (filtered graph, shorter path) Each filter is asked sequencially to process a path A filter consumes some parts of a path, and returns a filtered metadata graph Filter can delegate to a particular filter or delegate back to filter chain Output: Finally the filtered metadata graph is transformed to a directory view

20. Architecture of SemFS: Example for Filters Filter-List: [Favourite, Artist , Main] Main-Filter processes “\My Music“ and returns filtered graph, containing all music-resources. Delegates to Artist-Filter. Artist-Filter processes “\Fatboy Slim“ and returns all resource, that have e.g. the dc:creator „Fatboy Slim“. Delegates to filter chain. Favourites-Filter processes “\favourite“ and returns all resources accesses more than n times in the last x days Main Artist Favourite \My Music\Fatboy Slim\favourite \Fatboy Slim\favourite \favourite

21. Architecture of SemFS: Example for Filters Filter-List: [Favourite, Artist , Main] Main-Filter processes “\My Music“ and returns filtered graph, containing all music-resources. Delegates to Artist-Filter. Artist-Filter processes “\Fatboy Slim“ and returns all resource, that have the rdfs:label „Fatboy Slim“. Delegates to filter chain. Favourites-Filter processes “\favourite“ and returns all resources accesses more than n times in the last x days Main Artist Favourite \My Music\Fatboy Slim\favourite \Fatboy Slim\favourite \favourite View( path )  Metadata-Table Contained files, Contained directories, Metadata: getName, getDate, …

22. Architecture of SemFS: Class-Handlers Different content-types have different meta-data and different ways to read/write E.g. local bookmarks, remote bookmarks, images, mp3 files, … Idea: access each class of content types via a Class-Handler Metadata: getName, getDate, … Can be cached in metadata graph Updates Metadata: setName, setDate, add file, delete file, move file Binary+Metadata: read file, write file, trunc file Other metadata operations, e.g. setGenre in ID3-tags

23. Architecture of SemFS: Summary: Metadata Graph, Filters, Class-Handlers Metadata graph (RDF) Holds all metadata acts as a unifying cache, when content has ist own metadata Filters Create the view, organise resources (not limited to files) Class Handlers Handle different content types and manages binary data

24. Using SemFS: A tagging file system (TagFS) One Filter: Tag-Filter Consumes a single path segment a Contained files: Returns all resources tagged with a Contained folders: All tags b , for which the conjunctive query a+b is not empty One Class Handler Delegates classic metadata queries to underlying file store Binary content resides in a folder of a classic file system Tag-queries handled by metadata graph New problems  No two files may have the same name (when rendered) Rename at insertion or display time

25. Conclusions TagFS Many locaction property Browsing only until result is small enough, no further Orthogonality of information dimensions Query refinement Compatible with existing applications SemFS An efficient way to implement virtual file systems with pluggable semantics Allows rapid and solid prototype creation Interoperability with Semantic Desktop applications Download Prototype at http://isweb.uni-koblenz.de/Research (Linux only, WebDAV/Windows in progress) Thank you. Questions?

26. BACKUP

27. Contact Information Max Völkel (author, presenter) Forschungszentrum Informatik, Karlsruhe, Germany http://www.fzi.de Stephan Bloehdorn (author) Institute AIFB, University of Karlsruhe, Germany http://aifb.uni-karlsruhe.de / Olaf Görlitz (author) and Simon Schenk (author) ISWeb, University of Koblenz-Landau, Germany http://isweb.uni-koblenz.de /

28. Architecture

29. Tagging Ontology