Or?
I18n is about Internationalization = I(18 chars)n. Even the name by itself
should make it obvious that it's not about multilingual text (German:
Internationalisierung = I19g?). See the details at
www.w3.org/International/. I18n is about processing, not about storage
representation. Unicode is about multilingual storage (
www.unicode.org).
Much of I18n applies to Unicode storage, and issues related to choosing one
encoding over another (
www.w3.org/TR/i18n-html-tech-char/)
If you need to store Korean, you have to use some form of 2 byte
representation. DBCS is the older mechanism, Unicode the standardized
technique, both are supported by DB2 (DBCS is called GRAPHIC).
Issues of multilingual collation are linguistic, not representational.
Knowledge of the language is essential. German sorts ö equivalently to o,
whereas Swedish sorts it after z, and English writers tend to drop the
umlaut complete, leading to nonsensical pronunciation (schon and schön are
two different words with different meaning). Collation order is an
intractable problem: consider the problem of how to sort a table containing
both German and Swedish names - any choice you make will be totally wrong to
the other nationality, and hence you need to consider not the origin of the
data but the origin of the consumer! On the other hand, Swahili is correctly
sorted using the general Unicode sort.
You need to read up on both I18n and Unicode and understand what each is
about and how they complement each other. I've given you URLs above.
You also need to do some thinking about your problem, and understand its
ramifications. Getting hold of people with experience in other languages,
including complex scripts (Arabic, Chinese, Japanese, Korean, Thai) and
other alphabets (Hebrew, Russian), as well as differing rules when using
accented latin characters (most continental European, Vietnamese) will
undoubtedly help you understand the consequence of multilinguage databases.
You'll also learn that issues such as sorting (your "way slower" comment)
are in many ways the least of your problems. Fortunately, most systems
provide highly efficient processing of a range of common languages, but not
all languages. DB2 or SQL Server, for example, do not have native Swahili
support (your example); Windows XP introduced Swahili, Windows 2000 didn't
support it (other than through Unicode).