Re: LYNX-DEV minor display problem (?character 0xA2?)

[Drazen Kacar]

Klaus Weide wrote:

> A significant part of Lynx's audience is affected by the "problem" that,
> when Lynx Options are set in a certain way, Lynx will believe the user
> and act according to the Options setting.
> Now what is the way you suggest so that Lynx can detect, in a universally
> applicable way across all terminal, terminal emulators, and windowing
> systems, what display character set is in use and what translations may
> happen on the way between Lynx's output and what the user sees, so that we
> can get rid of the "display character set" field on the Options screen
> and avoid the "problem" that some users insist on setting it *wrong*?

There is, but it's just a theory for now. LC_* environment variables (on Unix)
should be set properly and Lynx could inspect them and act accordingly.
LC_CTYPE, in particular, should be set to display charset. The good thing
with this is that in some cases it can be set from /etc/login & co. files
on login, which means that user does not have to know a thing about it if
there is a competent admin around.

There are several bad things. First, recognized values for this variable
are not the same on all systems. For Latin 2 on Solaris it should be
LC_CTYPE=iso_8859_2 and on OSF it's LC_CTYPE=hr_HR.ISO8859-2 (or something
like that). This creates certain problems with telnet & co. since meaningless
value would be propagated to remote system.

Another problem is that most of the systems don't have tables for non Latin 1
code pages. The value of LC_CTYPE on Solaris, for example, is the name of
a file with ctype.h info for a particular code page (man chrtbl). That
file must be located in the specific directory and there is no way for
any user to override that. Solaris comes with iso_8859_1 table, but without
iso_8859_2. Those two tables differ in information about two characters
only (I think), BTW. Now, I need LC_CTYPE set because I want my editors
to work properly (delete word command, for example). I have it set to
iso_8859_1, because I know that table exist on every Solaris system
and it gets propagated via telnet, rlogin & ssh. If Lynx takes it for
my display charset, it would be wrong assumption, since it's usually Latin 2.

All in all, LC_CTYPE cannot be expected to contain reliable information.
Perhaps display charset menu can have LC_CTYPE option, which would mean
that Lynx will use LC_CTYPE value for display charset. This shouldn't be
default. Another approach is to allow LC_CTYPE autodetection from lynx.cfg,
which should be disabled by default. If some admin spent weeks in tuning
LC_CTYPE variables for his lusers and he knows it would never contain
the wrong value, he could enable autodetection in global config file.

> > I think all other chars in the 0x80..0x9f range can be displayed
> > directly.  Also, IBM PC chars in the range 0x01..0x1f can be displayed
> > with the "ESC [ = <decimal> g" sequence.  This might be useful...
> 
> Apparently a vendor specific hack.  Linux has other specific hacks to
> enable display of some glyphs in that range.  And so on.

And most of them won't work with screen, I suppose?

-- 
 .-.   .-.    Life is a sexually transmitted disease.
(_  \ /  _)
     |        address@hidden
     |        address@hidden
;
; To UNSUBSCRIBE:  Send a mail message to address@hidden
;                  with "unsubscribe lynx-dev" (without the
;                  quotation marks) on a line by itself.
;