Pangloss

By Steve Purkis

What is it?

A Web-based Multi-Lingual Terminology Management System
-- or --
A dictionary (in different languages)

Why is it being built?

• There are other systems out there
• The good ones are proprietary
  (fair enough, you say)
• But prices are artificially inflated.
• Expect to pay £15K+ for licenses
• Don't forget consulting fees...
• And other components...

So?

• Monetary barrier to localization companies
• A shame, because they're good products...

Who's building it?

• London Advice Services Alliance (LASA)
  Multikulti, community translations
• Web Architects
  Multikulti site
• quiup
  Pangloss for multikulti

Why Open Source?

• Open standards (TMX, XLIFF, ...) are emerging
• A common char. encoding (UTF-8)
• Means vendors can share data now
• proprietary file formats war ala MS .doc is coming to an end (I hope!)

Why Open Source?

• So OSS solutions can interoperate with proprietary ones
• an in-expensive alternative...

Enough with these delusions of grandeur...

• Pangloss is not quite ready to compete
• It needs to be finished first
  moreover...
• It's a terminology management system, not a translation memory tool

Moving right along...

So what does it do?
• Manage terms, of course!

In a nutshell:

• Categorize, translate, & proofread terms
• Search terms
• Build glossaries from web pages
  (well, not yet)
• Manage other things (users and such)

Lets have a look, shall we?

http://staging.quiup.com:8888/

So how was it done?

First, some history...

• Project began in March
• 1 full-time developer [mostly]
• 1 part-time htmler
• 2 whiteboard diagrams
• 1 notebook w/loads of scribbling

And surprisingly:

• A functional spec + a tech spec were produced
• Both still resemble the current product!
  (oooh...)

But:
• they don't go into too much detail (on purpose)
• it's not the biggest project in the world.
• it's easy to be consistent when there's only 1 developer

The Pretty Picture...

The Framework

• OpenFrame
  
  + OF::WebApp
  + Pipeline::Config
• Pixie

What, no database?

So I have to write some things myself
• Collections --> tables
• Searches --> selects
• Ordering search results

But I don't have to:
• do OO *and* data modeling
• write SQL
• write an abstraction layer

Other dependencies...

build_requires:

	'Test::More' => '0.01',
		

requires:

	'perl'			=> '5.8.0',
	'YAML'			=> '0.35',
	'Date::Format'		=> '2.0',
	'MIME::Types'		=> '1.06',
	'Error'			=> '0.15',
	'Petal'			=> '1.0',
	#'Module::Build'	=> '0.01', # soon
		

A lot of things went wrong...

• 1 person, too many jobs
  (new business, accounting, marketing, technical architect, front-end coder, back-end coder, sysadmin, ...)
• Juggling projects
• Lack of persistent user feedback
• Developer gold plating
• Over...
  ...Documented
  ...Budget
  ...Schedule (well, kinda)

But that's ok

• I've run other projects before...


• I know that's how you learn


• And besides...

A lot of things went right...

• project is well organized
• ~6,000 lines of code got written
• fairly easy to read
• 860+ tests
• design decisions were made
• API's emerged
• I'm happy with most of it...

A lot of things went right...

• a decent model emerged
  ie: business objects, collections, editors, search
• high-level configuration of controller
  ie: config.yaml
• separation of front/back end logic
  ie: term editor
• localization can be done in front end
  ie: error handling w/flags

Things are still going.

• the to-do list is long
• probably incomplete
• functionality not yet implemented
• system limitations unknown
• searches scalable?
• XLIFF, TMX, PO (ie: gettext) support
... etc ...
• TODO

If you're interested...

• Pangloss staging server:
  http://staging.quiup.com:8888/


• Pangloss mailing list:
  http://www.email-lists.org/mailman/listinfo/pangloss


• Pangloss code base:
  
  Not yet on CPAN
  http://staging.quiup.com/perl for now

That's it.

References

• LASA: http://www.lasa.org.uk/
• Web Architects: http://webarchitects.co.uk/
• Quiup: http://www.quiup.com/

Questions?

spurkis [at] quiup.com