Here is a brief comparison of CJK package and pLaTeX (HLaTeX for Korean should be similar):
First thing is that CJK is virtually never used in Japan. Hence style files specially tuned for (Japanese) local typesetting traditions are available only for pLaTeX. On the other hand, quite a few macro packages on CTAN like inputenc, fvrb-ex, etc. assume TeX compiler to be 8-bit clean and hence incompatible with pLaTeX. Needless to say, babel which relies on inputenc cannot be used with pLaTeX.
Nextly, DVI->PS drivers for standard LaTeX as well as for Omega/ Lambda does not support multibyte characters. In fact their support is restricted to Type1, Type3 and Type42 fonts which, Adobe has made it clear for ages, are legacy font formats not given a proper status as PostScript resources. IMNSHO it is a lazyness of Omega developers for not supporting CID fonts, at least Type9 fonts. Consequently PS/PDF files generated by CJK package and/or Lambda have non-standard single byte Type1 and/or Type3 Kanji fonts split from multibyte fonts embedded. On one hand this makes documents intolerable large, on the other this makes documents printable on systems without multibyte fonts installed. However xpdf cannot display embedded fonts in PDF documents. Naturally DVI->PS drivers for pLaTeX supports multibyte fonts but only for Japanese.
emacs+CJK seems a reasonable choice for multilingual purpose. But if you are ready to hack locale definition files, you can experiment CJK-LyX with JISX-0213 charset which contains all LTR Latin-X characters. At present this is quite hard since there are a few hundred Kanji characters in JISX-0213 which are not yet given Unicode codepoints and the scalable fonts backends of X does not yet support JISX-0213.
(Shigeru Miyata)