Warning: this page requires your system has some japanese fonts, such as Internet Explorer language pack or properly configured Mozilla/Netscape6 
NEC, Microsoft, etc decided they should include two sets of roman letters
in the common Japanese encodings. Ideally, these should be clearly different
because the standard ASCII set from 0x41 to 0x7a is exactly half the width
of the "double width" set, starting from 0xA3C1 in EUC-JP encoding. To a
typical Jap, however, those two look identical when it comes to entering
text or reading it. Example of double-width and standard width letters next
to each other:

£Á£Â£Ã£Ä£Å£Æ£Ç£È£É£Ê£Ë£Ì£Í£Ï£Ð£Ñ£Ò£Ó£Ô£Õ£Ö£×£Ø£Ù£Ú
ABCDEFGHIJKLMNOPQRSTUVWXYZ

Another huge problem is the so called "Japanese space". 0x20 in ASCII, and
0x2020 in EUC-JP, these elusive pieces of blank space are absolutely
indistinguishable to an average Jap.

Compare for yourself: "¡¡" vs " "

example taken from a e-mail message subject line:
  D  39 Apr 28 ENA      
  (1909) [tlinux-users-j:02526] KDE ¤È£Ç£Î£Ï£Í£Å¤Ë¤Ä¤¤¤Æ    
  notice how the user cannot tell the difference between half-width characters
  in "KDE" and "GNOME"

Note, shittily written search engines and CGI scripts will not make a 
difference between them, either. (For obvious reasons - you would have
to write code to turn full-width characters (different byte sequences) into
half-width, same with full-width spaces).