[thelist] Multi-language site infrastructure?

Administrative HQ english_offline at yahoo.com
Thu Mar 27 10:37:06 CST 2003


Hi. 

We run a multilingual site. I quickly glanced at the
archive Jeff pointed at.

I'm neither a programmer nor a developer but have
learned a number of things from experience and
provided assistance to the technical people who did
our site.

Some things to note:

All depends on the type (purpose) of the site. Is it
commercial? Just info?

Do you want search engines to find you according to
foreign language searches?

What languages are involved?

Our site (for the moment) is in English, Spanish and
Hebrew. It is a commercial site. Everything I have to
say is based on that.

Some things from our experience:

I must disagree with the points of view expressed that
*titles* can remain in English. If you want speakers
of various languages to find you, the titles must
contain the search words in each language that you
want people to find you with. Google picked up our
site in all three languages, but we were not getting
good results until we "customized" our titles AND
content. specifically, they use the words in the title
and in the HTML on the pages. This had special
ramifications for Hebrew (which would also apply to
Arabic and possibly others). Let me know if you need
to do that.

*Search engines* which work on other bases must be
given the exact URL for the language pages your are
promoting, and a description, etc. in that language.

For proper *display*, something like <HTML lang=x
dir=z> is not adequate. You must specify charset. We
do this with an include file.txt  specific to each
language.

To make *links* to various language pages visible to
all we use a graphic (text) which displays regardless
of the fonts, etc. the user may have.

Hebrew (or other right-to-left languages) require
special attention as the pages flow in the "wrong
direction" which can effect "simple" things like
"align=right".

If your site sends emails, autoresponses, etc. use the
latest versions of ASPEmail of Jmail. Earlier versions
do not support other languages well. Cdonts does not
work.

Plain text should be OK for Western European languages
provide you specify 8bit. For right-to-left languages,
plain text presents terrible problems in all spheres.
Punctuation, dollar signs, numbers can jump around,
appear reversed, etc. Use latest versions mentioned
above (don't forget to specify subject line encoding)
and hope. TEST A LOT BEFORE USING. We found that most
things remained correct in Hebrew using windows-1255,
while UTF and other encodings tried were a disaster.

Mixed language text is especially problematic. Don't
try to use an Arabic sentence containing English (e.g.
a URL). Use separarte lines.

HTML came out best for us in these cases using tables
where English text and non-English were in separate
td's.

We have not seen a server component which will send
multipart, so we send two emails, one plain text one
HTML. Not an ideal solution, but the folks at
ASPemail, Jmail Cdonts and others need to get used to
the fact that the Web is international.

Our private list manager and server can do this, but
must be integrated with the ASP pages on the site
which is a problem.

*Translation* should never be done by machine or an
amateur. Use a native speaker. Languages are not just
words. They include entire cultures, modes of address
and styles. What sounds just fine in English could
come out down right uncultured in Spanish. Languages
like Japanese are much more "dangerous". Rely on an
experienced native's intuition.

*Consider results*. Do you allow user feedback,
contact? If so, expect people to write to you in their
native language and be sure you have a way to handle
that. Can you reply? Do you need someone "on call" to
read and respond to such emails?

A few thoughts.

Good luck.
David








--- Peter Loron <peterl at standingwave.org> wrote:
> I'm looking into making some of the sites I create
> have support for 
> multiple languages. Something similar to the
> capability at:
> 
> http://news.bbc.co.uk
> 
> Any references or example code people can point me
> at? Specifically, 
> I'm looking for ideas/best practices, etc on how to
> architect the pages 
> and backend code (I typically use PHP) so that I can
> minimize the 
> duplication of effort for each language. Thanks!
> 
> -Pete
> 
> -- 
> * * Please support the community that supports you. 
> * *
> http://evolt.org/help_support_evolt/
> 
> For unsubscribe and other options, including the Tip
> Harvester 
> and archives of thelist go to:
> http://lists.evolt.org 
> Workers of the Web, evolt ! 


__________________________________________________
Do you Yahoo!?
Yahoo! Platinum - Watch CBS' NCAA March Madness, live on your desktop!
http://platinum.yahoo.com


More information about the thelist mailing list