Re: lynx-dev Re: 283dev15 for Win32

[Henry Nelson]

> +#if defined (WRITE_ON_TOP_OF_SCREEN)
>      move(0, 0);
>      clrtoeol();
>  #if defined(CJK_EX) && defined(SH_EX)
>      addstr(str_kcode(last_kcode));
>  #endif
> +#endif /* write on top of screen */

Doesn't make sense to me.  What's the difference between "#if defined(CJK_EX)
&& defined(SH_EX)" and "#if defined (WRITE_ON_TOP_OF_SCREEN)"?  You're not
going to want the character set labeled unless you're reading Japanese
(CJK_EX), and may not want it cluttering up the screen anyway (SH_EX).
  
> > They should be independent, neither within the other.
                                ^^^^^^^^^^^^^^^^^^^^^^^^
This statement is *wrong*.  Sorry.

> So, without CJK extensions, I don't get the functions I mentioned.
> It would be nice to have descriptions of these functions so we'd

They are described on Hiroyuki's page, in Japanese.  I assume anyone
who wants CJK would be able to read Japanese, on the further assumption
that little or no testing of these extensions is happening with the other
two languages (remotely) involved, Chinese and Korean.

> know how they are used without much digging.  I would guess the
> "k" in str_code means Katakana?

Or _kanji_?  The reference, I believe, is to whether the page is in
sjis or euc-jp.  I got a few URLs from Hiroyuki to look at, but haven't
had the time to really investigate what the problem is.  Usually I use
Lynx in "screen", but it's not safe to do testing in that environment
(which slows the testing process).  Good thing Fote isn't around, otherwise
I'd get chastized again (maybe Klaus will anyway) for not understanding
what's going on.  My, probably mistaken, impression is that ASSUME_CHARSET
has no meaning for CJK, and that before Hiroyuki's work, the code set of the
page was automatically (thus the "AUTO" at the top of the screen is the
default behavior) guessed by the Sato/Asada functions.  The problem is that
they never could handle the 7-bit kana and occasionally messed up on some
of the multi-byte characters as well.

Recent MSIE and Netscape do a 99.999% correct job, so there must be some
way to determine the correct code, even if interspersed and incorrectly
labled.  Hiroyuki tells me that he has given up trying to do that kind
of guess work, and finds it easier, *and more reliable*, to manually specify
the code set for "problem pages," thus he did the modifications.

__Henry