There are various definitions of the term Software Engineering. A concise definition can be found in IEEE [1993]: 1.) the applicationof a systematic, disciplined, quantifiable
approach to the development, operation, andmaintenance of software; that is, the applicationof engineering to software, and 2.) the study of approaches
as in (1).
Global Software and Localisation Testing is a key aspect ofIBM’s Global Software Engineering division, and is a professioncommonly referred to as GVT (Global VerificationTesting). At the IBM Dublin 留學生dissertation網Software Lab GVT is a disciplinethat underpins the success of many IBM products that arereleased across many global markets and is therefore animportant and necessary activity for global success. It is thereforean established and integral part of IBM’s end-to-end softwaredevelopment process.
The key motivation for GVT is the common knowledge thatit is difficult to sell computer hardware/software which doesnot support the prevailing language and national customs.Therefore IBM and most software development organisationsnow produce software that is tuned to the needs of local marketswith the notion that early success is essential, andrespecting standards is a smooth prerequisite to ensuring customersatisfaction and quality. The IBM GVT process attemptsto ensure that standards are uniformly implemented andrespected across all of their software applications, as consistency
in approach and methodology not only aligns processacross development teams, but also delivers a consistentexperience to the end users.Interestingly, one of the first attempts to establish a standard
took place in the year 1120 (at the time of the 1stCrusades). The pioneer was King Henry 1st of England. Henrymandated that the ell (the ancient yard) should be the samelength as his arm and he insisted that this distance would bethe standard unit throughout his lands. The ell (45” in length)was used until recently for measuring cloth. Henry’s counterpartacross the channel must have been of smaller stature, asthe Flemish ell is 27 inches or 3/4 of a yard… Standards,Standards, Standards!!!
STANDARDS IN GLOBAL SOFTWARE
ENGINEERING AND TEST
Standards in software engineering and test are critical torepeatability and incremental success. There are many internationalstandards organisations that have helped guide theevolution of International standards and best practices in theglobal software industry. These organisations have membersfrom a number of large and small organisations, and serve asa “police force” to ensure that the introduction of new technologies,protocols and standards are monitored carefullyagainst global requirements. Everyone is a vested communitymember of this police force and adherence is key!The end-to-end process of softwareglobalisation and localisationis complex as Fig.1 shows. Starting with the earlystages of development right through the process of release#p#分頁標題#e#
and manufacturing there are many potential pitfalls that canresult in mistakes or fundamental errors. Several checkpointsneed to exist in the process to allow engineers to identifyproblems early so as to motivate timely solutions.Standards exist right from code inception (where developersneed to design code to be culturally, platform, and linguisticallyneutral) right through to when the final CD is cutTHIS ARTICLE WAS FIRST PUBLISHED IN LOCALISATION FOCUS VOLUME 3 ISSUE 3 SEPTEMBER 2004 63
ENGINEERING GLOBAL
SOFTWARE – A SYSTEM TEST
STANDARDS PERSPECTIVE
Pat O’ Sullivan & Maria Hyland(where formatting and writing characteristics also need torespect established standards).
To help industry in the difficult task of comprehending andrespecting the various global standards we note that there aremany international standards organisations who have helped
guide the evolution of international standards and best practicesin the software industry. These organisations, which havemembers from a great number of both large and small organisations,
serve as our “police force” to ensure that the introductionof new technologies, protocols and standards is monitoredcarefully against international requirements.
The International Standards Organization (ISO) is a consortiumwhich was set up to encourage the interchange of
character data between countries which do not share commoncharacter sets or alphabets. ISO's main focus is in recommendingprotocols & GUI rendering mechanisms to preserveand present data in an accurate and persistent way. ISOhas made great efforts to evolve legacy 7-bit protocols to 8-bitso that data above 127 decimal can be transported correctly.
This effort is most noticeable in the TCP/IP protocol. ISO hasalso defined the widely accepted 8-bit character set table,allowing for international languages to be included. Examplesinclude ISO-8859-1 for US and Western European countries,IISO-8859-2 for Eastern European countries, ISO-8859-3 forCyrillic countries, and so on. ISO has also evolved the Unicodestandard, and this is a superset code page and character setrepresentation which captures all possible characters for allpossible countries and code pages. As we know, ISO’s workhas influenced many aspects of Global Software Testing.
The American National Standards Institute (ANSI) is anational standards body working on standardisation of alltypes. Noticeable efforts include the development of the ANSIcharacter set and the standardisation of all of the variouscomputer programming languages, particularly C and C++.The National Institute of Standards and Technology(NIST) is a member of ANSI and represents the United Statesgovernment. NIST advises the US government on standardsregarding hardware and software publishing. A good examplehere is the exporting of encryption technology outside of theUS, where it was http://www.400190.com/dissertation_sample/jisuanjiwangluo/#p#分頁標題#e#once illegal to make the 128 bit encryptiontechnology (used in the language applications like LotusNotes) available to countries outside the U.S.The Institute of Electrical and Electronic Engineers
(IEEE) was formerly concerned with hardware standardisation.More recently the organisation has become much moresoftware focused. The organisation has also started to look at
more fundamental engineering standards such as interfacedefinition, software integrity and interfacing of systems. Such
standards feed various aspects of Design, Development andGlobal Software Testing.
The European Computer Manufacturers’ Association(ECMA) is dedicated to standardisation of hardware-orientedfacets of computers. It is concerned with the standardisationof paper sizes, page layouts, conversion routines and devicecompatibility.Global software testing needs to consider language specific
challenges and users' disparate needs. The Unicode projectevolved to address this and involves a number of companieswhich include Apple Computers, Xerox, Claris, IBM,Microsoft, Sun Microsystems, and many more. The key objectiveof the consortium is the delivery of a multi-byte characterset to permit many countries to share and exchange data in away which maintains the persistency of data across the variousoperating systems and protocols. Other objectives include a)the research, development and standardisation of code pointsfor the characters and character sets of new or emergingcountries; b) the implementation of specifications to describethese new character tables in terms of decimal values whichcan be referenced in software; and c) the influencing of datainterchange standards at many levels.LISA was established to promote standardisation in technicalterminology by organising technical seminars, forums, andso on. The goal of the forum is to agree on standard terminology.Today LISA is working to standardise thousands of termsfor the information technology industry. It is also attemptingto raise the overall recognition and image of the localisationindustry. There are several companies on the LISA technicalsteering board including IBM, Apple, Alpnet, Berlitz, DEC,IBM European Language Services, ICL/UK, INK International,Mendez, Microsoft, NCR, Oracle, Rank Xerox, Sequent,SoftTrans International and many more.Since its establishment in 1995 at University CollegeDublin, the Localisation Research Centre (LRC) has developedfrom a funded three year project into a world-recognised
http://www.400190.com/dissertation_sample/jisuanjiwangluo/research centre with a strong industrial backing manifestedin its high-profile Industrial Advisory Board. The
centre has representatives from academia and both overseas
and indigenous companies, and has become a respected
representative body for the industry in Ireland. Its annual
conference has become one of the major industry events
attracting interest from the globalisation community nationally#p#分頁標題#e#
and internationally. Key areas to the LRC’s credit
include the coordination of the efforts of the localisation
industry, the provision of a forum for information dissemination
and sharing, liaising with other relevant interest
groups, and the pursuit of excellence in research, development
and information dissemination. Since March 1999 the
LRC has moved to its new designated laboratories at the
University of Limerick in Ireland and is continually expanding
its own services and research.
64 L O C A L I Z A T I O N R E A D E R 2 0 0 4 - 2 0 0 5
Software Bugs Software
Bugs
Linguistic / Translation
Bugs
Tested
Software /
Product
Refs
Refs
Refs
Analysis
Data
Manufacture
Product
Bugs
Bugs
Bugs
Fully
Tested
Product
Build
Scripts
Test Scripts
Source Code
& Bug Fixes
Translated Strings
Translated
Files
Translated
Product
Files
Prepared for
Translation
Analysed
Product
Source Code
Test Scripts
Preparation
Material
Product
Developers
Bug Database
Product References
Linguistic References
Translator References
L18N Databases
Bug Database
Bug Database
Develop
& Test
Product
Analyse
Product
Manufacture
& Release
Prepare
for
Product
Translation
Quality
Engineering
Localise
and Build
Product
Translate
HCI
Product
Builders
QE
Engineers
S/W
Engineers
Translators
Product
Beta/QE/Testers
Product
Analysts
Product
Translators
Product
Linguistics
Customer
Bug Databases
L18N Databases
Build DBs
1
2
3
5
4
6
7
Fig. 1: Software globalisation and localisation process
The Localization Institute is an organisation providing
quality resources, training, seminars and conferences on
localisation, internationalisation and international business
development. Their mission is to enable companies to succeed
in international markets by providing focused training
for professionals and by promoting the sharing of experience
and information among them. The Localization Institute is
perhaps best known for its annual “Localisation Roundtable”
which is a gathering of localisation leaders. The Localization
Institute has also conducted a number of seminars such as
“Introduction to Localisation” as well as two seminars on
“Writing and Designing for Localisation”.
The International Workshop on Internationalisation of#p#分頁標題#e#
Products and Systems (IWIPS) attempts to bring together
internationalisation and localisation specialists from industry,
academia and government. It was founded in 1998 by
Girish Prabhu and Elisa del Galdo. IWIPS is small by design,
to maximise networking and promote a shared experience
among delegates.
The standards organisations serve to promote best practice
and provide an education forum and reference point for their
members. As Scott McNealy, Chief Executive Officer of Sun
Microsystems, pointed out, standards arise from common
usage and best applicability to the specific problem domain.
The standards organisations usually have a very large number
of members from worldwide development organisations, and
they in turn agree and implement the recommended standards
to which smaller companies need to abide to remain
compatible and to survive. Without standards, international
incompatibility within programming languages, systems, protocols
and ultimately geographical regions would hinder
progress and best practice.
AN IBM GLOBAL SYSTEM
TEST PERSPECTIVE
Global verification testing at IBM's Dublin Software Lab
includes NLS (National Language Support) and aspects of TVT
(Translation Verification Testing). GVT is therefore the
process whereby all aspects of Software is verified for functionality
and suitability in world markets. IBM’s Dublin
Software Lab’s System Test teams see GVT as a key part of our
overall testing paradigm. Under the lab's GVT umbrella IBM
plan and execute several classifications of complex tests on
English, European and Asian configurations. One aspect to
this testing is the need to ensure that all aspects of the software
are adhere to recognised industry standards.
Significantly, IBM and most other reputed software development
organisations discourage proprietary/break-away standards,
and view break-away approaches as bad practice that
leads to unforeseen downstream problems. Not having an
established and commonly agreed methodology ultimately
results in customer problems in aspects of Hardware and
Software implementation. During IBM System Test engineering
standards are pervasive in cross platform protocol/interoperability
testing including some fundamental aspects of LDAP
(e.g. to respect the InetOrg Directory Schema), HTTP/HTTPS,
SSO, SSL, SOAP, SIP, WSDL, GSM/CDMA/GPRS, and so on.
IBM’s need to test on a multitude of platforms (AS400, AIX,
Linux SuSe+Redhat, S390, Windows 95/98/2000/
XP/ME/2003) and client browsers (Internet Explorer,
Netscape, Mozilla) aggregated with the need to support the
incremental product versions creates a potentially exponential
test matrix. However, ensuring that standards are respected
on one operating system platform often permits the ability to#p#分頁標題#e#
scale testing on the other platforms. This is essential for containing
testing times and costs, however to reduce the matrix
one needs to understand aspects of the underlying standard
and its implementation which can oftentimes be difficult.
There are other key dimensions to testing that standards
are correctly respected. For example, testing in one, two and
three tiered client/server systems will surface unique challenges
when compared to testing on simple isolated end user
configurations. Also, interface testing of global systems needs
to be conducted to ensure seamlessness of data. Cross platform
testing also presents unique challenges as each platform
has its own nuances. Character encoding decisions may be
different, and may generate conversion issues when going
across platforms. Also, unique aspects of configuration paramaterised
by locale and character set nuances can give rise to
new GVT concerns. Implementation specifics and run time
behaviours can also give rise to unique issues – for example,
client side cellular devices may behave differently and implement
standards in an inconsistent way.
Some other less obvious testing concerns where standards
apply include the fact that automation needs to be written
language neutral, and platform neutral; concerns for Bi-
Directionality (RTL/LTR) need to be respected; IMEs (Input
Method Editors) need to be tested in the native language and
platform; graphics and icons need to be verified as culturally
suitable and neutral; currency formats and time formats need
to be cognisicant of the local requirements; and so on.
Perhaps less obvious but implied standards in global software
testing include sacred symbols that need to be respected in
software and tested for (where appropriate). Examples here
include the Christian Cross in Christianity, the Menorah in
Judaism and the Crescent in Islam. Unlucky numbers are also
interesting, for example 2 and 514 in traditional Chinese
belief; 4 is an unlucky number in Hong Kong, Korea and
Taiwan; 4 in China and Japan suggests death. The number 7 is
unlucky in East and West Africa while the number 13 is
unlucky in most countries. Problematic gestures in
Graphics/Icons is also worth noting, for example blinking the
eye in Hong Kong and Taiwan; the folded arms in Fiji and
Finland; the “Stop” gesture in Greece and Nigeria; and the use
of the left hand to point in Islamic cultures.
Industry typically expects conformity to standards from the
bigger players, and to serve as role models. However this is
not always seen in practice. Hence a significant bias is placed
on testing in areas such as sorting and sort sequences,
searching, date/time, authentication, data interchange and
integration, decimal separators, locales, and DBCS (Double#p#分頁標題#e#
Byte Character Set). In the area of character sets it is worth
pointing out that testing for China’s GB18030 character set is
a mandated requirement from the PRC government if software
vendors wish to release to that market.
L O C A L I Z A T I O N R E A D E R 2004 -2005 65
From a development perspective some obvious concerns
can have major impacts on the testing of global software, and
can give rise to many questions and testing implications that
are underpinned by inherent and established best practice
standards:
 Are all executable program files designed so that they do
not need to be rebuilt after translation?
 Is all information relating to text strings, dialogues,
menus, messages, coordinates, sizes, fonts, etc., completely
separated from the code?
 Is all translatable information stored in platform's standard
resource file format, so that localisation tools can function
without modification?
 Does all user-visible text correctly display all permutations
and combinations of foreign characters, both single-byte
and multi-byte?
 Is full multi-byte support provided for all user-visible
text?
 Is it possible to use accented characters as hit-keys or
accelerator keys in dialogs, menus, and lists?
 Are all string comparison algorithms implemented to
take into account the requirements of languages other than
English?
 Are all date, time, currency, list separator, measurement,
number formats, etc., fully resourced for localisation?
 Is all help text separate from the program’s core executable?
 Are there any assumptions made concerning the sizes
and availability of fonts or font information?
 Are all aspects of the product designed to be culturally
neutral?
 Do all resource files contain appropriate comments and
documentation?
 Is it verified that any resourced text which is dynamically
constructed (such as error messages) can be translated
appropriately, and that variable parameters can be rearranged
without corrupting the string?
 Does the UI (User Interface) design allow for text strings
to expand by up to 30% when they are translated?
 Are all icons, cursors, and bitmaps culturally neutral
and designed so that any text is overlaid at runtime?
CONCLUSION
In recent years developments have materialised to help
bridge data interchange problems along established and well
debated standards. Examples include XLIFF, CSS, XML, XSL,
SRX, OLIF, DXLT and TBX just to name a few. It is also encouraging
to see progress in automation tools, from respected
software vendors such as Rational and Segue, that allow for
automation scripts to be written in a way that allows them to
be reused on language versions of software. From a translation#p#分頁標題#e#
perspective encouraging evolution in Translation
Memory, Machine Translation and Translation Web Services
are bridging fundamental gaps that existed in the past.
IBM have learned that it is one challenge to respect standards,
but how the standard is implemented may not be properly
standardised. Issues have been observed in the IBM
Dublin Software Lab where 3rd party products read and display
IBM formatted HTML/HDML/WML correctly, but there
have been problems when attempting to render in reverse.
Some software vendors embed proprietary tags, and parse for
these during run-time execution effectively bypassing agreed
standards under the umbrella of proprietary logic. This presents
problems for “good citizens” who try to interpret the
incoming data based on the established standards. The test
harness subsequently becomes “test the standards” + “test
the exceptions”. Too many times teams have exerted engineering
and testing efforts to manage the exceptions and
oftentimes containing test plans and test cases is impossible as
they spiral out of control.
Critical issues in the rapidly changing global marketplace
challenge the formal international standards system. The time
to complete an international standard is often out of sync with
the marketplace's needs. In the information world web years
are measured in months. Development, publication and adoption
of international standards must keep pace and this is one
key problem. International standards enjoy global acceptance
and worldwide agreement on content and acceptance, as
opposed to consortia and proprietary international standards.
Any attempt to derail these strengths devalue and trivialise the
meaning of international standards. 
Pat O'Sullivan is a Test Architect at IBM's
Dublin Software Lab. He completed his BSc
in 1991, his MSc in 1997 and his PhD in
2001, and both of his IBM sponsored MSc
and PhD HCI research projects won industry
awards. He can be reached at patosullivan@
ie.ibm.com
Maria Hyland is a Development Manager
at IBM and has many years (15) experience
in the field of Software Engineering, Software
http://www.400190.com/dissertation_sample/jisuanjiwangluo/Development and Globalisation. Currently
Maria is managing the Dublin and India
System Test teams for one of our key projects
- Lotus Workplace. She can be reached at
[email protected]
66 L O C A L I Z A T I O N R E A D E R 2 0 0 4 - 2 0 0 5