Re: [cjk] output with source han and xdvipdfmx

[Werner LEMBERG]

> 1. If you are doing new documents, and do not have a lot of baggages
>    and opinion about how Latex *should* work, LuaTeX and XeTeX can
>    both access arbitrary platform fonts.  I have no experience with
>    LuaTeX; I tried XeTeX, and found the change in layout,
>    word-breaks, hyphenations unfamiliar.

I've tried both, and thanks to the uniform `fontspec' interface, you
usually don't have to worry about the underlying engine.

>    [...] if you are writing a new document in LaTeX and needs to do
>    a lot of languages or a lot of different fonts, those are the
>    ways to go for flexible handling of fonts, and other typographic
>    features of non-English.

Exactly.

> 2a. The main difference between going through postscript with dvips
>     ->ps2pdf vs directly to pdf via pdfTeX/dvipdfmx is graphics.  If
>     you need psfrag or any of the latex packages which depends on
>     running through ghostscript, then you have to use the former.

As mentioned in a previous mail, `pstool' might be an alternative so
that you can use modern engines.

> 2b. There are a number of differences between pdfTeX and dvipdfmx -

In general, dvipdfmx produces *much* more compact PDFs.

>     the former has a larger (more English-speaking) community, and
>     the latter was/is run by a smaller number of Korean people.

This is no longer correct: dvipdfmx is now maintained directly in
TeXLive (essentially by all TeXLive developers).  Note that dvipdfmx
is also the base for `xdvipdfmx', the DVI driver for xetex.  For this
reason, many glitches found in xdvipdfmx are backported to dvipdfmx.

>     One of the explicit goals of the latter is preserving
>     text-searchability, i.e., pdf's from dvipdfmx preserves encoding
>     information, and you can extract non-english text, cut-and-paste
>     from it (the same applies to the modern XeTeX/LuaTeX output
>     also), have its content indexed by a Google's search engine,
>     etc.  Whereas non-English dvips/ps2pdf and pdfTeX generated
>     pdf's loses encoding information and loses the full
>     compatability of being indexed by a search-engine.

This is only partially correct.  There exists the `cmap' LaTeX
package, which creates proper `CMap' PDF objects to make text using
encodings like T1 or T2A searchable.  Additionally, IIRC, pdftex
itself has got better support for the creation of `ToUnicode' mappings
a few years ago.

> 5. I'll like to make all of them work again, so I'll possibly try to
>    get dvips and pdfTex to work also; personally I am leaning on
>    dvipdfmx as it lets me use the font as is (sharing with
>    platform-viewing) without going whole-sale the XeTeX/LuaTeX way.

Thanks for working on this!


    Werner