Banal Because Format Checking is So Trite

1
BanalBecause Format Checking is So Trite

• Geoffrey M. Voelker
• University of California, San Diego
• Workshop on Organizing Workshops, Conferences,
  and Symposia for Computer Systems (WOWCS08)

2
This Talk is Not Very Interesting

• Banal is a format checker for PDF documents
• Deduces how a document was formatted
• Optionally compares it with a specification
• Intended for conference management systems
• Now being used in HotCRP and EDAS
• Seemed timely to document its genesis and
  implementation

3
Why?

• Preserving reviewer anonymity
• Acrobat javascript that calls home when pdf is
  loaded
• Assisting conference management tasks
• Ensuring anonymity rules
• Possibly helping do initial assignments by mining
  the bib
• Fairness
• Everyone else obeyed the rules
• Time
• Already enough time spent on reviewing
• Frustrated that abuse meant taking even more of
  my time

4
How?

• Convert PDF
• To XML (with pdftohtml)
• Track the locations of all segments of text,
  essentially form bounding boxes
• Compute margins, columns, body font, etc.
• Heuristics for page s, headers, footers, etc.

5
Where?

• A handful of SIGOPS/SIGCOMM conferences
• OSDI06, SIGCOMM07, SIGCOMM08
• Eddie Kohler has integrated it into HotCRP
• Henning Schulzrinne also integrated banal with
  EDAS
• Since 2006, used for over 800 events

6
So?

• What are our community goals for having
  formatting requirements?
• Evil Annoying trifles that negatively impact our
  ability to communicate our results and ideas?
• Helpful Reflect practicalities of publishing
  costs and community time?
• Not surprisingly, Im in the practical camp