lynx-dev Re: bookmarks encoding

[Ilya Zakharevich]

Leonid Pauzner wrote:

    > The display charset is a "random variable"; it is determined by the
    > current page charset on the architectures which support a display
    > charset switch.

    1998-11-18 (2.8.2dev.4)
    * implement EXP_CHARTRANS_AUTOSWITCH for OS/2 EMX (Sergey Svishchev).

Actually, the code of EXP_CHARTRANS_AUTOSWITCH is marked as "Linux
only" in the sources; I do not think it did anything useful until I
added another branch: CAN_SWITCH_DISPLAY_CHARSET.  The actual
"hardware" switch is done only on OS/2 (but it should be trival to do
on other architectures supporting this switch too).

    > And I do not see *why* this flag is needed.  When newly created, a
    > bookmark file is written in a correct encoding (I mean the first
    > bookmark in the file is correct).  When subsequent bookmarks are
    > written, lynx can read the bookmark file, deduce its encoding, then
    > write the bookmark (using Numeric Character References or not -
    > depending on the encoding).

    Just imagine: you read cyrillic texts and european texts (latin with
    ascents), and the first bookmarked document is in iso-latin-1.

I should not imagine this.  I have many bookmarked files like this.

    In this case all subsequent cyrillic links will be transliterated,

Why transliterated?  Just NCR-ed.

    which is not nice.

Comparing to what?  The current behaviour is to show garbage (without
NCR_IN_BOOKMARKS), and have *always* uneditable cyrillic (with
NCR_IN_BOOKMARKS).  The proposed behaviour will switch to NCR in the
problematic situations only.

    You suggest a new kind of "randomness", dependence
    on the charset of the first bookmarked document.

What is new with this, it is the default behaviour now, right?  If you
use mixed-display charsets, and want editability, you *must* have
bookmark charset to be variable.

Ilya

; To UNSUBSCRIBE: Send "unsubscribe lynx-dev" to address@hidden