libwww - The W3C Protocol Library

1
libwww -The W3C Protocol Library

• Großes Schwerpunktseminar WIUniversity of
  Applied Sciences Gießen-Friedberg
• Stefan Sabatzki

2
Contents

• Introduction
• Structure libwww
• Programming with libwww
• Conclusion

3
Contents

• Introduction
• What is libwww?
• Why libwww?
• Structure libwww
• Programming with libwww
• Conclusion

4
What is libwww?

• Generic framework for building web applications
• Written in C
• Pluggable modularity
• Means to provide most common Internet access
  methods
• Transmit data in many different media formats
• Dataflow to and from the server

5
What is libwww? (2)

• First version implemented 1992 by Tim Berners-Lee
• Development at CERN
• 1994 libwww moved from CERN to W3C
• 1998 released as opensource
• As of September 2003 W3C stopped work on libwww
• As of January 2004 libwww officially belongs to
  the Open Source Community

6
Why libwww?

• Experimenting and prototyping
• Performance, modularity and extensibility
• Free and open source code
• Mailing lists and active community

7
Contents

• Introduction
• Structure libwww
• Design Model
• Request/Response Paradigm
• Data Flow
• Threads, Eventloops and Filters
• Modules as Statemachines
• Programming with libwww
• Conclusion

8
Design Model

• Layering as design model

9
Design Model (2)

• More demonstrative

10
Request/Response Paradigm

• Application issues request
• Libwww fulfills request
• Presented to application on arrival
• Simultaneous requests handled by Librarycore

11
Data Flow

• Streams are used to transport data
• Derived from generic stream
• Protocol streams
• Converters
• Presenters
• I/O streams
• Basic streams

12
Data Flow (2)

• Structured streams
• Derived from generic stream
• Accepts structured document
• Ordered tree-structured arrangement of data
• Each instance is associated with SMGL parser
• Each instance is associated with corresponding
  DTD

13
Data Flow (3)

• Cascaded streams
• Stream chains
• Setup before data arrives

14
Data Flow (4)

• Setup after data arrives

15
Threads, Eventloops and Filters

• Not thread-save
• Implements pseudo-thread model
• Uses non-blocking sockets
• Based on callback functions
• Before/After-Filter
• Global and local filters
• Registered at runtime

16
Threads, Eventloops and Filters (2)
17
Modules as Statemachines

• Since libwww 3.0
• Protocol modules implemented as statemachines
• Part of thread-model
• Keep track of current state in communication
  interface

18
Modules as Statemachines (2)

19
Contents

• Introduction
• Structure libwww
• Programming with libwww
• C Simulation
• APIs and Library Interfaces
• Simple Example
• More Complex Example
• Conclusion

20
C Simulation

• Construction/destruction
• _new / _delete (HTRequest_new /
  HTRequest_delete)
• Data hiding
• Inheritance
• Explicit pointer casting
• PRIVATE, PUBLIC Makros

21
APIs and Library Interfaces

• Set of APIs called packages
• Win32 DLLs
• Unix separate static libraries
• Package interface exported via single include
  file WWW.h
• Some important packages
• Basic Utility Packages
• Core Packages
• Initialization Packages
• Transport Packages
• Protocol Packages
• Parser Packages

22
Simple Example

• Displays all links in document
• Applicable to text, html/xml tags, etc.//
  snippet
• ...
• HText_registerLinkCallback(foundLink) .
• HTEventList_loop(request) ... foundLink (...)
  
• HTAnchor dest HTAnchor_followMainLink(...)
  
• char address HTAnchor_address(dest)
  HTPrint("Found link s\'\n", address)
• HT_FREE(address)

23
More Complex Example

• Rudimentary commandline browser
• See project www.dsw

24
Contents

• Introduction
• Structure libwww
• Programming with libwww
• Conclusion
• Whats missing?
• Facts about libwww
• Personal Opinon

25
Whats missing?

• Not thread-safe
• No cookie-jar, only parsing/generation
• Consistent usage of RegEx
• C representation

26
Facts about libwww

• Who uses libwww? No one?
• Sample applications on project homepage
• No reviews, benchmarks, comparisons
• Not bug free
• Competitors (mostly UNIX)
• WinInet
• Libghttp
• Libcurl
• Libhttp
• Neon

27
Personal Opinion

• Typical opensource project
• Tricky installation
• Feels old lt gt IS old
• Desperate attempt to reach OOP
• Non-trivial usage, but very flexible and potent

28
Thank you for your attention

• ?