d internationalisation
standards, and current practices in open source localisation. Section
3 presents our main result, a canonical mapping of the PO format to
XLIFF. We then look at how this format can be incorporated into
existing build systems and development processes (Section 4).
Section 5 builds on these foundations, discussing the need for other
exchange standards in open source localisation. We conclude, in
Section 6, by summarising the findings of this paper, and by examining
the future of OSS localisation.
Adopting Standards-based XML File
Formats in Open Source Localisation
By Asgeir Frimannsson and
James M. Hogan,
School of Software Engineering
and Data Communications,
Queensland University of Technology,
Brisbane, Australia.
In recent years, major localisation vendors and key standards organisations have agreed on open XML-based standards
for storage and exchange of data in the localisation process: The Translation Memory eXchange (TMX) file format for
exchanging translation memory data, the TermBase eXchange (TBX) format for terminology exchange, and the XML
Localisation Interchange File Format (XLIFF) for extracting and storing locale-dependent resources in a common file
format. Up until recently, very few open source tools have supported these formats, and hence very few open source
projects have adopted them in the localisation process. A majority of open source applications continue to use Gettext
and the Portable Object (PO) file format. This paper evaluates the case for adopting XLIFF in localisation processes
currently based on the Gettext toolkit, and examines the usefulness of other standards such as TMX, TBX and
Translation Web Services (TWS) in these processes.
James M. Hogan Asgeir Frimannsson
10 LOCALISATION FOCUS RESEARCH.loc DECEMBER 2005
Research
2 Background
2.1 Open Source Localisation
The term ‘open source’ relates to the practice of freely sharing access
to the source code of a product, allowing anyone to extend or modify
a piece of software. The open source development process has
been described as a user-driven, just-in-time approach, driven by a
global developer community (Berglund and Priestly, 2001).
Development is driven by demand for the product within this community,
and new features are implemented as a result of requests
from the user base. Raymond (2000) describes the process as a
‘bazaar’ where software is released early and often, the software
process is open and transparent and work is delegated as much as
possible. In contrast, traditional software is developed by closed
teams having long release cycles and is only released after being
thoroughly tested. Surveys reveal that fewer than one in five open
source participants are paid for their involvement (Hars and Ou,
2001). These characteristics imply that localisation of open source
software is an ongoing process, driven primarily by voluntary
efforts, in response to constantly changing source code.
Localisation of software messages is handled in most open source
projects by Gettext — as specified in the OpenI18N Globalisation
Specification (Free Standards Group, 2003). The Gettext library is
based around two file formats:
• The Portable Object (PO) file format: a simple string table for
storing translation units in the localisation process
• The Machine Object (MO) file format: a binary representation
of a string table — used by an application to retrieve translated
strings at runtime
Whilst the
