Page ThesisDPPXVIIBerichte

Weekly reports for thesis DPPXVII

pre

weeks (1.2 - 12.2.10)

• getting to know Saros
  • Eclipse and Plug-in background (reviewed)
  • the videos, theses from C.Jacob, M.Rintsch and S.Ziller
  • Code view
    • event flow
    • understanding code rules and conventions
    • applied patterns
  • Threading
  • Jupiter and synchronization

• background to whiteboards
  • checking existing whitboards (Groupboard, Coccinella, ReLaTe + see beyond)
  • drawing, writing and mindmaps
  • how to implement a whiteboard
    • understanding Xpairtise and SWT example whiteboard
    • SVG

• how to integrate it
  • Saros' activity structure
  • whiteboard implementation structure

• knowledge (overview)
  • Test-first and test-driven development
  • UX-Design, prototyping and iterative programming/testing

week (15.2 - 19.2.10)

• SVG as proper internal representation
  • informing about libraries Batik and SVG Salamander
  • read documentation of Batik and its uses (i.e. GMF)
  • writing code snippets
    • finishing a pencil whiteboard as Eclipse View using Batik (no Saros yet)

• first plan for diploma timetable

weekly reports

week 1 (4 days) (22.02.-26.02)

Activities

• Release manager
• prepare presentation
• understanding Batik and SVG
• how far is standardization respective SVG + XMPP
• general planning of the thesis

Literature

• Batik documentation
• SVG documentation
• http://xmpp.org/extensions/xep-0113.html
• http://xmpp.org/extensions/inbox/whiteboard.html and whiteboard2.html

Problems

• ill (wisdom tooth)

week 2 (actually 2 days, however, a lot of work at home) (01.03.-05.03)

Activities

• understanding Batik
  • proper usage of UpdateManager and SVGGraphics2D
  • distinguish between different possibilities to draw
    • create DOM elements our own
    • usage of SVGGraphicsD2
    • usage of SVGShape
  • batiks basic features: zoom, scroll and rotate
    • integration
• add a vertical SWT Toolbar to snippet
• fixing bug #2958625
  • learn about Saros invitation structure
• fixing user feedback when peer cancels
  • Eclipse Jobs usage

Literature

• Batik documentation
• SWT documentation
• finally after asking to many questions Eclipse Jobs API

Problems

• ill, much worse (wisdom tooth)
• fighting with SWT layouts

week 3 (5 days) (08.03.-12.03)

Activities

• add Tools structure to snippet
  • support of batik's built-in zoom and pan
• Performance testing with batik
  • Usage of SVGGraphics2D, SVGShape or manual creation of XML Elements
  • Reason for AWT-Event-Queue delay (too many straight lines while using the pencil)
    • try out to draw pencil on Graphics object first, to XML on mouse release only
• Shared XML Editing (SXE) as most supported SVG + XMPP approach
  • protoXEP only; but experimentally used by the IMs Gajim, Psi and Miranda (obsolete)
  • search mailing lists to follow the acutal state of standardization and known issues
    • jabber Standards, proto-xep, Shared Editing and Gajim develop

Literature

• http://xmpp.org/extensions/inbox/sxe.html
• http://xmpp.org/extensions/xep-0228.html
• Gajim and Psi documentation

week 4 (3 days) (15.03.-19.03)

Activities

• add IActivity and Saros manager infrastructure for distributed whiteboarding
  • distributed editing works!
• contact initiators of Shared XML Editing protoXEP
  • Joonas Govenius is not working on it anymore
    • in his opinion SXE might be a little bit too verbose
  • no response from Peter Saint-Andre yet
  • current implementations like psi and gajim differ somehow
• try to contact psi's and gajim's developers list (no response yet)
• how to synchronize SWT scrollbars with batik

Literature

• batik, SWT and Saros documentation

Problems

• wisdom tooth extraction
• surprised how verbose it is to add an IActivity
• it does not seem to be easily possible to uncouple the whiteboard from Saros

weeks 5-9 (22.03.-21.04)

• virus, probably mononucleosis

weeks 9-10 (6 days) (22.04.-30.04)

• reintroduction to my work, catching up with current state of Saros
• Responses for SXE
  • Psi uses an outdated version (no jingle for session negotiation)
    • no time/interest to improve this
    • that's why incompatible with Gajim which uses latest version
  • found a new contact who tries to improve SXE and who would be interested to try check compatibility
    • spare responses
• SXE not complete for our purpose yet (selection, blocking etc. messages)
• trying to fix batik rendering problems, especially while resizing
  • even the simplest whiteboard has these problems. Response from batik mailing list: these are SWT Swing integration issues and they cannot help
  • entering deep to batik rendering leads only to underlaying rendering engines:
    • with Direct3D rendering cheesy and doubled painting occures, unusable
    • with openGL the picture shakes verically while resizing
    • in both cases batik rendering might get stucked while resizing, workarounds lead to ugly rendering/flickering and perform very bad
  • no response yet form the gef-batik project

• start work on networklayer to change Socks5 as first choice and add IBB fallback
  • studying Hennings smackx extension (good work!

Problems

• still not completely recovered

weeks 11 (6 days) (02.05.-07.05)

• still no responses for batik problems or SXE improvement and implementation
  • more batik problems appeared respective redirecting events to eclipse (Discussion http://old.nabble.com/SVG-editor-in-Eclipse-td26056864.html), not yet improved
  • considering using GEF again

• GEF (Graphical Editing Framework)
  • reintroducing myself to GEF (used to work with it)
  • creating a simple whiteboard view, figure based, however, rectangular selection of freeform shapes
    • GEF is the better solution, offers command stack, properties management, selection, doesn't have the problems of SWT integration. Batik offers a lot but difficult to use within eclipse (a bit hacky).
  • Main problem: proper free-hand drawing
    • introducing to Eclipse Sketch project (free-hand drawing pattern regocnition)
    • published 28.04. only, would be a nice extension for any whiteboard for programmers

• Batik
  • contacted Gerrit who used to work with GEF and batik and created the SVG composition eclipse plugin
    • http://developer.berlios.de/projects/svgcompost
      • batik rendering on graphical contexts (using GMF adapter), above transparent rectangular GEF figures (bounds)
      • no freeform shapes, no drawing, a littly bit buggy
    • advises to use GEF for batik because of GEF's editing modell
      • Problem: DOM API is unhandy as modell in MVC pattern
  • in the first moment SVG and batik is deferred
    • if time, investigating to use DOM as GEFs modell representation
    • waiting for answers for SXE improvement

• ECF
  • I got familiar with the container and provider concept, some tutorial
  • offers a kind of multi-IM integrated in eclipse
  • views extendible by additional features
  • XMPP support (without jingle)
    • for jingle there exists a provider as part of a VoIP provider (GSoC project)
      • http://wiki.eclipse.org/VoIP_via_the_ECF_Call_API_and_the_Jingle_Protocol
      • jingle: 15 classes, offers abstraction
    • MUC support
  • it wouldn't take too much effort to write own providers for jingle, SOCKS5 and IBB
  • However, XMPPConnection in Saros has more than 80 references, needs more investigation to estimate the effort if we ever want to switch to ECF

• networt refactoring
  • introducing to Saros network layer
  • adapting SOCKS5 and IBB to Transport-Conneciton concept
  • helping Henning to find deadlock cause in IBB

Literature

• ECF documentation, GEF documentation and tutorials, smacks API

weeks 12 (3 days) (10.05.-14.05)

• debugging and refactoring the network layer
  • improve smacks API changes (what is missing?)
  • transport hierarchy
  • getting SOCKS5 to work with IBB fallback
  • study synchronization and take care of properly releasing resources
  • improving mediated SOCKS5 connections if server only allows unidirectional connections
    • wrapping two unidirectional connections in a bidirectional one if necessary

Literature

• smacks documentation, OpenFire documentation
• XEP 65 SOCKS5
• XEP 47 IBB

Problems

• ejabberd, saros-con.imp.fu-berlin.de, was configured wrong
• getting to know the APIs and network layer
  • i.e. how to configure OpenFire or smacks properly

weeks 13 (5 days) (17.05.-21.05)

• refactoring Networklayer
• testing and fixing SOCKS5
• updating preference page to new network layer implementation
  • deeper understanding of SOCKS5 mechanics and distinguishing necessary preferences
  • applying preferences on reconnect
  • change DataTransferManager to enable "force file transfer by chat"

Problems

• understanding static SMACK behavior
• deeper understanding of SOCKS5 mechanics lead to some doubled work

weeks 14 (5 days) (24.05.-28.05)

• finishing to implement preference page and preference mechanics
• some more refactoring and renaming in the network layer to avoid confusion
• removing unused classes and obsolete methods
• change to always try to establish a direct SOCKS5 connection
  • establishing a connection from both side and discard one (if not needed for wrapping)
• testing
• find problems when using IBB
  • dispose bytestreams in DISCONNECTING state
• studying proper locking of observer editor (Bug 2971277)
  • drag and drop
    • solution: SWT filter
  • copy and paste
    • possible solutions:
      • deactivate these commands (does not work, get reset)
      • disable the editor (does not work, gets reset)
    • writing to eclipse platform mailing list, responses are pending

Problems

• reimplementing done work because of late understanding of the best practice respective SMACK
• editor preferences get reset without obvious reason
• we are locking the editor just by intercepting keys

Literature

• SWT, eclipse documentation and newsletter

weeks 15 (5 days) (31.05.-04.06)

• more work and tests with the network layer
  • added simultaneous connection establishment from both sides to quickly achieve a working (direct) connection although establishing from one side only wouldn't work
  • completing documentation and debug output
• GEF
  • solution for continuous freehand drawing
    • studying Sketch API again
      • no option for us: painting on GC would get painted over by concurrent edits
    • writing to GEF newsletter
      • no response
    • Studying GEFs code there results a proper solution: usage of feedback layer
      • hacking existing code lead to a working result
      • a proper implementation consists of adding specialized tool, edit policy, request and command classes for continuous drawing
        • we might distribute this to GEF and Sketch afterward to achieve better compatibility
  • testing for concurrency
    • concurrent creation of new figures seems to work without graphical problems
    • pending: concurrent modifications and consideration of proper lock-mechanisms

Literature

• learning more about Java concurrency for proper simultaneous connection establishment
• GEF code and API
• GEF newsletter
• Sketch API code

weeks 16 (6 days) (07.06.-12.06)

• improving network layer again
  • catching exceptions if one connection establishment does not work, the other one can be used
  • testing and trying to find out reasons for Ubuntu invitation delay and missing direct SOCKS5 connections
• considering model representation (this and next week):
  • preliminary: stay SVG conform, for saving, loading sake and to develop a protocol for distributed XML editing
  • Using prebuilt library classes and generate DOM only for saving
    • SWT and draw2d
      • hide concrete shape in drawing method, platform dependent (no independent XML editing protocol)
    • AWT shapes
      • platform independent representation and easier than DOM
      • batik could be used to generate SVG from AWT shapes
        • unfortunately AWT shapes do not maintain hierarchies
      • however, transforming shapes to SVG and back didn't result very efficient
    • custom implementation
      • still the SVG has to be mapped and generated
        • just to export to SVG, GMF and batik could be used, no loading
      • reinventing the wheel
  • Using DOM
    • platform independent representation
      • model might kept independent form concrete DOM implementation
    • mapping
      • searching a working SVG DTD
      • using JAXB to generate JAVA classes
      • very verbose, less flexible than plain DOM
      • shared SVG editing only instead of shared XML editing
    • plain DOM
      • any XML will do it
      • flexible
      • best standardization approach might be to use DOM Level 2 Event Model
    • generate infrastructure using model-based software engineering
      • details:
        • generating a Ecore Model from SVG DTD
        • using openArchitectureWare to generate infrastructure
        • only editor and mapping between draw2d figures has to be implemented
      • all advantages of plain DOM
      • reduces the effort to implement the verbose GEF infrastructure
      • easier to stay SVG compatible (hierarchy is given by Ecore Model)

Literature

• JAXB
• JAXP
• DOM
• EMF

Problems

• Understanding complex DOM API
• finding an error free SVG DTD

weeks 17 (5 days) (14.06.-18.06)

• considering model representation (continue)
• creating working GEF prototype without distribution
  • postponing concrete model representation
    • still need to know more about DOM, XML-Java Mapping and code generation, especially the DOM Event Model
    • model prototype to evaluate different approaches:
      • wrapping the DOM in Java beans using property supports and listeners
  • implementing GEF infrastructure, copy, paste, undo and redo
    • not yet working completely
  • adding rectangle tool
  • implementing freehand drawing tool, edit policy, commands
  • presenting current state

Literature

• GEF

weeks 18 (3 days) (21.06.-23.06)

• Fixing GEF editor and implementing left over work
  • copy, paste, undo and redo works as well as zoom
• start to add collaborative working feature to model
  • consider event flow

Problems

• using GEF editor infrastructure in ViewPart
  • i.e. static zoom combo box, retarget actions, toolbar and selection issues

weeks 19 (3 days) (30.06.-02.07)

• GEF editor
  • fixing pencil tool feedback with zoom and scroll,
  • documenting, restructuring

• fixing network bugs:
  • 3023491
  • 3020135
  • 3023695
  • 3023576
  • Stacktrace when establishing SOCKS5 bytestream connection

Problems

• 2 days sick, furthermore flatemate moved out from one day to another, had to buy new fritch and washer

weeks 20 (6 days) (04.07.-09.07)

• GEF Editor
  • finish collaborative work model prototype
    • adding Saros infrastructure
    • WhiteboardManager as connection to Saros
  • adding ellipse tool, default sizes on drag'n'drop, pencil drop added (for convenience)
  • cleaning whiteboard project
    • removing tryouts and unused libraries like batik

• Testing
  • fixing several small issues when working distributed
    • cycling events
    • doubled creation
    • fix for getElementById() issues
  • reducing the amount of events fired

• Studying DOM Level 2 event model
  • a proper collaborative XML editing framework and communication protocol should be based on this event model
    • offers standard naming for DOM mutation
    • also selection and focus events
    • preferable to SXE approach

Literature

• DOM API

Problems

• Several DOM API issues like getElementById() in Java that doesn't always work like expected
• interaction between undo, model and commands i.e. delete, undo, redo (mixed up order, pending) that doesn't allow using GEF API classes

Comments