Flat ODF: the under-estimated flavour of Open Document

Transcript

1. Fridrich Štrba, Software Engineer
   Flat ODF:
   the underestimated flavour
   of Open Document
   

   View Slide

2. Whois?
    Software Engineer in SUSE Linux
   Enterprise
    Used to work for SUSE on LibreOffice
   and OpenOffice
    Diverse background
    FLOSS enthusiast
    Working in free time on various
   projects including LibreOffice
    Document Liberation Project
   

   View Slide

3. What is Flat ODF
   

   View Slide

4. “Normal” ODF File
    Package OpenDocument File
    Chapter 3.1.3 of ODF 1.2 specifications
    Collection of xml files in a ZIP container
   
   
   
   
   
   
   
   
    Images and OLE objects sub-streams in the ZIP container
    Referenced by xlink:href attribute
   

   View Slide

5. How it looks inside
   47 b- stor 14-May-05 22:23 mimetype
   26087 b- stor 14-May-05 22:23 Thumbnails/thumbnail.png
   1526 bl defN 14-May-05 22:23 meta.xml
   11315 bl defN 14-May-05 22:23 settings.xml
   44472 bl defN 14-May-05 22:23 content.xml
   37148 b- stor 14-May-05 22:23 Pictures/100002010000038D00000226C198A9BD.png
   99784 bl defN 14-May-05 22:23 Pictures/100185C800006D7C00004242BCCE6537.svg
   385 b- stor 14-May-05 22:23 Pictures/100002010000001E0000001E57F6E610.png
   181031 bl defN 14-May-05 22:23 styles.xml
   0 b- stor 14-May-05 22:23 Configurations2/popupmenu/
   0 b- stor 14-May-05 22:23 Configurations2/menubar/
   0 b- stor 14-May-05 22:23 Configurations2/images/Bitmaps/
   0 b- stor 14-May-05 22:23 Configurations2/toolpanel/
   0 b- stor 14-May-05 22:23 Configurations2/progressbar/
   0 b- stor 14-May-05 22:23 Configurations2/statusbar/
   0 bl defN 14-May-05 22:23 Configurations2/accelerator/current.xml
   0 b- stor 14-May-05 22:23 Configurations2/floater/
   0 b- stor 14-May-05 22:23 Configurations2/toolbar/
   1367 bl defN 14-May-05 22:23 META-INF/manifest.xml
   

   View Slide

6. “Flat” ODF file
    Single OpenDocument XML Files
    Chapter 3.1.2 of ODF 1.2 specifications
    Contains the whole document
   
   root element
   
   office:mimetype and office:version attributes
    Images and OLE objects inlined
   
   
   [base64 data]
   
   

   View Slide

7. ZIP Storage (1)
    Main entry point is the Central Directory End located at the end of
   the ZIP file.
    Scanning for it towards the end of the file
    Whole file must be present
    Pointing at the Central Directory containing entries
    For name access, iterate over them
    Need to seek back to read the entry
    Which contains pointer to Local File Header
    Seek to local file-header offset
    Need to seek back again
    The content of the stream comes after the header
   

   View Slide

8. ZIP Storage (2)
   FILE
   E NTRY 1
   
   FILE
   E NTRY 2
   
   FILE
   E NTRY 3
   
   FILE
   E NTRY 4
   
   Local header 1
   Local header 2
   Local header 3
   Local header n
   R elative offset 2
   R elative offset 1
   R elative offset 3
   R elative offset n
   File entry 1
   File entry 2
   File entry 3
   n
   C E NTR A L
   D IR E C TO RY
   

   View Slide

9. Advantages of Flat ODF
   

   View Slide

10. No Need for Special tools
     No need for compression and stream extraction tools
     ODF documents can be generated manually
     Developer modifying a document
     XSLT tools
     Programmatic generation of documents
     Further processing of documents
     Easier parsing of document and extraction of relevant information
    

    View Slide

11. Sequential Access to File
     Difference with ZIP file
     No need to know the end of the file to start to parse it
     Exchange file-format par excellence
     Possibility to communicate document as SAX messages
     Communicate XML to an application sequentially
     Possibility to stream document over wire
     Collaboration protocol?
     Communication with CMS systems?
    

    View Slide

12. Simple API for XML (SAX)
     A class containing typical functions
    
    startDocument();
    
    startElement(const char* name,
    std::map &attributes);
    
    characters(const char* characters);
    
    endElement(const char* name);
    
    endDocument();
     An XML producer calls functions of the class
     Passing to it a content in abstract XML form
     An XML consumer is inheriting from the class
     Processing the content received from producer
    

    View Slide

13. Issues with Flat ODF
    

    View Slide

14. Duplication of embedded binaries
     Package ODF file
     The embedded images and ole objects are in a special storage
     Referenced by link
     Possibility to reference the same file several times and save space
     Flat ODF file
     Embedded binary objects are in-lined as base64 data
     Inlined on every reference
     Potentially huge files in corner cases.
     Possible solution in specification extension
     Data section listing binaries as we do with gradients or bitmap fills
     Needs “political” will and pass through specification process.
    

    View Slide

15. Bitrot and underspecification
     Few ODF producers support it out of the box
     LibreOffice is supporting this file-format out of the box
     New features disregard it
     Unspecified how some features in package transform to flat
     Risk of implementation specific solutions
     Risk of abandon and bitrot
     Possible solution in specification extension
     Need some extra work and “political” will
    

    View Slide

16. Examples of Flat ODF use
    

    View Slide

17. LibreOffice UNO Filter API
     XSLT based filters
     com.sun.star.comp.Writer.XMLOasisImporter
     pushes to LibreOffice flat ODT
     com.sun.star.comp.Calc.XMLOasisExporter
     receives from LibreOffice flat ODS
     XML based filters
     com::sun::star::xml::sax::XDocumentHandler
     com.sun.star.document.ImportFilter pushes SAX messages to
    com::sun::star::xml::sax::XDocumentHandler
     com.sun.star.document.ExportFilter is itself a
    com::sun::star::xml::sax::XDocumentHandler and receives SAX messages
    from LibreOffice
    

    View Slide

18. File importers
     libodfgen generators
     OdtGenerator – generates ODT from librevenge::RVNGTextInterface
     Used by libwpd, libwps, libmwaw, libabw, libetonyek, libe-book,...
     OdgGenerator – generates ODG from
    librevenge::RVNGDrawingInterface
     Used by libwpg, libvisio, libcdr, libmspub, libpagemaker, …
     OdpGenerator – generates ODP from
    librevenge::RVNGPresentationInterface
     Added for the use in libetonyek
     OdsGenerator – generates ODS from
    librevenge::RVNGSpreadsheetInterface
     Added for needs of libwps and libmwaw
    

    View Slide

19. Thank you!
    www.documentliberation.org
    

    View Slide