This is the outline for the CodeCon presentations on FeedParser I'm giving in February.

Outline

• Introduction
  
  • Based on the NewsMonster parser infrastructure (XSLT)
  • Designed for use within Rojo (an online RSS aggregator)
  • Event based not DOM based
  • Jakarta Commons
  • Apache 2.0 Open Source License
    
• Challenges with building a feed parser
  
  • Too many standards
    • RSS (0.9, 0.91, 0.92, 1.0, 2.0)
    • Atom (0.3-0.5 and all draft specs (IETF work in progress))
    • OPML
    • FOAF
    • Changes.xml
    • RDF
    • XFN
    • HTML (link parsing, relations, nofollow, meta tags, generators, etc)
    • Modules (dc, aggregation, content, etc)
      
  • Semantic confusion:
    • rss:entry vs atom:item
    • title issues across specifications (dc, rss, atom, etc)
      
  • Encoding issues
    • Invalid entity references
    • XML prefix prior to <?xml?> (usually XML comments)
      
  • Date encoding issues:
    
    • RFC822 (RSS 2.0)
    • ISO8601 (RSS 1.0 and Atom)
      
• Feed Event Model
  
  • SAX model
    
  • DOM on top (in the future)
    
  • SAX is about 12x faster
    
  • FeedParserListener:
    • init()
    • onChannel( state, title, link description ): void
    • onItem( state, title, link description ): void
    • onItemEnd(): void
      
  • General API not wire API
    
• HTTP issues (network API):
  
  • Timeouts
  • ETags (If-None-Modified)
  • If-Modified-Since
  • UserAgent
  • Correct string support via Content-Type
    
• Problems with DOM models:
  
  • Namespace matching doesn't line up correctly.
  • Doesn't (easily) support ad-hoc schema updates with extensions
    
  • Plugin API to pass events with vendor specific interfaces.
    
    • mod_bigcompany
    • BigCompanyEventListener
    • Totally isolated development.
    • Just register a SAX DefaultHandler to intercept your own events
      
• Autodiscovery
  
  • FeedLocator API
  • Atom + RSS autodiscovery support
  • Feed location via href
  • URL fishing (disabled by default)
    
• Blog Profiles
  • Flicker doesn't support HEAD
  • Invalid autodiscovery implementations
  • Avoid URL fishing
  • Profile discovery support
    
• Feed Creation
  • Same API can be used to create RSS feeds
    
• API
  • FeedParsing:
    
    • FeedParserListener (rss, atom)
      
  • Directory Parsing
    
    • FeedDirectoryParserListener (opml, foaf, changes.xml)

• • Content Parsing
    
  • Tag Parsing
    
• Thanks
  • Brad Neuberg
  • Rojo Team!