Knowledge Transfer

1
18th International Unicode Conference
Documentum and UTF-8 Converting Content
Management Software Product Line to Unicode
27 April 2001 Donald Ziff
2
Agenda

• What is Documentum?
• Documentums I18N Problem
• How Unicode UTF-8 Saved the Day
• Other Success Factors
• Demo

Documentum Proprietary and Confidential
3
About Documentum

• Documentum NASDAQ DCTM
• The Leader in Web and Enterprise Content
  Management Solutions
• gt 128M in revenue 1999. gt 800 employees.
• Over 900 Global 2000 customers with strong
  vertical focus
• Over 25 Offices in 10 countries

4
DCTMs I18N Problem

• Everyone agrees we need I18N to fuel growth
  especially in Asia
• Asian-certified product much more important than
  multi-lingual
• Although demand for multi-lingual is growing
• So why not I18N?

5
I18N Perception Problems

• Too Difficult wont fit into a development
  cycle
• Too much Overhead multiplies QA and Support
• Not Sexy no new functionality
• Lets look at these problems

6
I18N is too difficult

• Product Layers
• Server (built on RDBMS Verity)
• DMCL Client Library (C)
• DFC Foundation Classes (Java)
• DTC Desktop Client Win32 end-user client
• WDK Web Development Kit
• RightSite Legacy Web-Server Integration
• Web Publisher Web Content Management App
• Legacy clients Workspace (Win32), Intranet

7
History Lesson

• Server v3.1.6.INT, created by consultants for
  Japanese market, was expensive and time-consuming
• 3.1.6.INT attempted to internationalize all the
  layers in the DCTM architecture at once
• 4.0 was released without I18N changes
• 4.1 followed, the deltas from 3.1.6 to 3.1.6.INT
  became hard to apply

8
I18N requires too much overhead

• The DCTM server requires pharmaceutical-strength
  certification
• Dimensions of certifications
• 3 RDBMS platforms Oracle, Sybase, SQL-Server
• 4 Server OSs NT, Solaris, HPUX, AIX
• The 3.1.6.INT architecture introduced new
  dimensions, leading us to

9
Certification Hell!

• New certification dimensions
• 5 DCTM Server code-pages
• 5 RDBMS code-pages
• Market requires another dimension
• 5 Server OS Localizations
• 125 new times 12 old ? 1500 certs!
• Exaggeration, of course But still

10
I18N not sexy

• DCTM is a growth company, needs sizzle as well as
  steak
• I18N grows markets, but doesnt add much to
  marketing message
• To be fair new functionality is not just sexy
  it is essential to DCTMs continued survival
• Other priorities will move to the top

11
DCTMs I18N Requirements

• Crucial need support Asia from the main
  code-line. One binary for the world
• Backward compatibility essential
• Multi-lingual features would be a side-benefit.
  High on the wish list for a few key customers
• I18N project must be scoped down to be achievable

12
How UTF-8 Saved the Day

• UTF-8 moves safely through the server because
  anything that looks like ASCII actually is
• Standardizing on UTF-8 as the only supported
  internal code-page cuts down certification matrix

13
Lessons from Double-Byte Experiments

• EUC-KR 4.1 server works (basically)
• SJIS problems! double-byte characters whose
  second bytes are ASCII \
• Lessons
• Non-ASCII moves through the server safely
• String handling need not be double-byte aware, if
  ASCII always means ASCII
• Solution UTF-8!

14
UTF-8 ASCII is ASCII

• No need for special string handling
• Server 3.1.6.INT replaced all standard c string
  handling with calls to 3rd-party library
• With UTF-8, we stick with standard yacc and
  other legacy tools work fine
• Greatly improved perception (and reality) of how
  difficult I18N would be
• Now, its relatively low-impact

15
Its UTF-8, dummy!

• Use UTF-8 everywhere, cut down on certification
  dimensions
• Provides safe character-handling for Asia
• Even though multi-lingual is not a requirement
• Easier to support

16
Other Success Factors

• Rely on RDBMS services to translate between RDBMS
  code-page and UTF-8
• Market research cut back on OS localization
  constraints
• Transcoding infrastructure

17
RDBMS transcodes to/from UTF-8

• Oracle and Sybase transcode automatically SQL
  Server is a problem
• No need for new transcoding calls between Server
  and RDBMS lower impact
• Upgrade customers have non-unicode RDBMS no
  need for them to convert
• One less certification dimension!

18
Cut back on Localized OS certs

• Limit RDBMS for Asia for 4.2, just Oracle
• Localized OS certification not necessary for
  Europe

19
Transcoding Infrastructure

• Server must be aware of interface code-pages
• Transcoding done at the interfaces
• 3rd party transcoding used Uniscapes GlobalC

20
New I18N Architecture
Intranet Client
Administrator
Web Publisher
Custom WebApp
Desktop Client
WDK (Unicode)
Rightsite(NCS)
WorkSpace
DFC (Unicode)
ARP(NCS)
Web Cache
( UTF8) DMCL (4.2)
DMCL 4.1 (NCS)
e-Content Server (UTF8)
Legend
National Character Set
Unicode
File System
Verity
RDBMS (Unicode)
21
Demo

• Demo multilingual WDK
• If theres time, a quick look at localized
  Desktop Client (Win32 Client)

22
Conclusion

• UTF-8 was a crucial technology in DCTMs I18N
  strategy
• Provided an easy path for legacy C
• Supported specific Asian languages consistently,
  minimizing certifications
• Prepared infrastructure for multi-lingual
  requirements