Eiman Al-Awadhi asked:
> Do you know of any work that has been done to establish Arabic MARC as a
standard for Arabic cataloging?

The principal types of MARC are MARC 21 (from the Library of Congress) and
UNIMARC (from IFLA). Both MARC 21 and UNIMARC are implementations of ISO
2709, described in the UNIMARC Manual as "an international standard that
specifies the structure of records containing bibliographic data." In both
MARC 21 and UNIMARC, the structural elements (leader/record label,
directory, tags, etc.) are in ASCII, which is equivalent to the
International Reference Version of ISO/IEC 646. Although ASCII contains the
letters of the English alphabet, it is used in computing applications as an
international writing system.

The URLs cited by Ian M. Johnson describe a particular implementation of
MARC 21 for Arabic script.  Another MARC 21 implementation of Arabic script
was released by RLG in 1991. RLIN21 is used by leading Middle Eastern
libraries in the US and UK for their Arabic script cataloging; thanks to
their efforts, the RLG Union Catalog contains over 150,000 records with
Arabic script data.

The official specifications of MARC 21 are on the Library of Congress site
at http://www.loc.gov/marc/   In the MARC 21 documentation, you should pay
particular attention to the MARC 21 Specifications for Character Sets,
Record Structure, and Exchange Media
(http://www.loc.gov/marc/specifications/spechome.html) and the discussion
on the structure of multiscript records in the MARC 21 Bibliographic Format
(http://www.loc.gov/marc/bibliographic/ecbdmulti.html)

The official specifications of UNIMARC are on the IFLA site at
http://www.ifla.org/VI/3/p1996-1/sec-uni.htm  For the character sets
specified for UNIMARC, see Field 100 General processing Data.

Note that the data in MARC records is in logical order, FROM FIRST TO LAST.
The issue of left-to-right versus right-to-left is a DISPLAY issue. Both
MARC 21 and UNIMARC include a method to indicate that a field should be
displayed from right-to-left. Left-to-right is the default display order
for both MARC 21 and UNIMARC.

Particular uses of MARC 21 and UNIMARC occur in a particular language
environment, and this is where the issue of script arises. Both MARCs
provide for an initial default character set, and the character set that is
chosen as the default will depend on the language environment. (Some
countries have specified their own adaptations of MARC 21 or UNIMARC, but
this makes the international exchange of data more complex.)

Both UNIMARC and MARC 21 allow multiple scripts to be included in regular
fields (in MARC 21, this is designated as "Model B"). MARC 21 also
specifies "Model A", which consists of a basic record supplemented by
"alternate graphic representation" in linked 880 fields.

Both MARC 21 and UNIMARC allow use of Unicode instead of individual
character sets (however, the UNIMARC specification for use of Unicode is
inadequate). In MARC 21, either Model B or Model A can be used with
Unicode. The model chosen for implementation depends on whether there is a
need to have a transliterated record with the source script(s) in
supplementary fields (Model A) or whether multiscript data in regular
fields is preferred (Model B).

In the Arabic-speaking environment, Model B would be the logical choice,
and Unicode should be used instead of individual character sets (since this
is the trend in the IT world).

John Eilts makes the point (in paragraphs 2 and 3 of his posting)  that the
real issues for Arabic are with regard to cataloging standards, not with
regard to MARC. This is correct. MARC is merely a container, and is neutral
with respect to script and language. It is the *environment* where MARC is
used (what I have called the "locale" when writing about multiscript
issues) that determines which language and script is of primary importance.
For a discussion of "locale-specific" access points, see my IFLA Conference
paper "Multilingual and Multiscript Issues in Cataloguing"
http://www.ifla.org/IV/ifla62/62-alij.htm

The content of locale-specific access points is dictated by the cataloging
rules and whatever "sources of authority" are being used for name headings
and subject headings. These are the authoritative sources for cataloging
that will be language oriented (and so script oriented).

Joan Aliprand
Senior Analyst, RLG
2029 Stierlin Court, Suite 100, Mountain View, CA  94043
Phone: +1 (650) 691-2258
Fax: +1 (650) 964-1461
RLG's Web Site: www.rlg.org