Problem is not unique.  Faced by all information providers

Old approach -- SGML

• Information all stored in common format
• Same information produced in different formats
• originally designed for legal materials (IBM -- GML in 1970's)
• Only caught on with big publishers (Lexis, DOD, most text book publishers) because complicated and expensive.
• leading application program $15,000 per seat

SGML
 

• Core idea.  Separate materials into two parts. Control with two functions.
• First part ---  information
• Most frequently text but can be anything
• Second part -- information about information
• "Joseph Conrad"  on book binding of "Heart of Darkness" is information but also is information about information
• he is author
• it is part of binding
• should be emphasized
• show information about information with tags:
• <author> Joseph Conrad</author>
• <emphasis><author>Joseph Conrad</author></emphasis>
• <binding><emphasis>Heart of Darkness<author>Joseph Conrad</author></emphasis></binding>
• First Control Function-- Document Type Definition
• defines tags and rules for using them ( "no section 2 without a section 1")
• industry specific-- developing DTD is moderately difficult. Standard DTD's important
• Second Control Function-- Style Sheet
• defines how tags are presented ("for emphasis use purple and make it blink")
• specific to output device (may be printing press, maybe browser, maybe Palm Pilot, may be voice, etc.)