German mirror provided by / Deutscher Spiegelserver von  &  SEO LEO

PuTTY wish unicode-mappings

Home | FAQ | Feedback | Licence | Updates | Mirrors | Keys | Links | Team
Download: Stable · Snapshot | Docs | Privacy | Changes | Wishlist

It's well known that Red Hat 8 boxes start up in UTF-8 by default, and hence that PuTTY must be set into UTF-8 mode or else commands such as man(1) will Do The Wrong Thing.

I (SGT) have just had a chance to actually play with PuTTY connecting to a RH8 box, and discovered that even with UTF-8 enabled, you don't always get the right results. man(1) handles hyphen/minus characters inconsistently: in some situations it produces U+2212 MINUS SIGN, but in others it produces U+2010 HYPHEN. ("man rm" is a good way to show up both types.) Many Windows fonts (Lucida Console, Courier New) support the former but not the latter, so that even in UTF-8 mode there are nasty "unknown character" rectangles appearing in the man page.

I think a neat solution to this would be to have a list of compatibility character translations, mapping a single Unicode code point to a different single Unicode code point, so that U+2010 could be displayed as U+2212 (or even U+002D!). This feature would also be useful for displaying U+0060 and U+0027 as U+2018 and U+2019, to ease the conflict between TeX-era Unix users who quote `like this' and pedantic fonts in which those quotes utterly fail to match).

Points to consider:

• In its raw form this feature could be more pain than it was worth: anyone switching between fonts would have to add or remove compatibility mappings to reflect the available characters in the new font. I believe it should be possible (Character Map can certainly do it) for PuTTY to query the font at run time and determine which characters are unsupported, and to enable only those mappings necessary. That way, we can ship with a large number of default mappings which should Just Work for almost all fonts.
• Of course, this will stop the feature being useful for the TeX quotes problem! To preserve this aspect, we would have to be able to configure each individual mapping to be unconditional (applied no matter what), or conditional (applied only if the current font does not contain the source character).
• It seems clear to me that these mappings should be applied at the display level, not the terminal input level: the terminal data structures should store the original untransformed characters, so that cut and paste doesn't depend on the selected font. (But, aargh, what about RTF pasting which specifies the font?! Perhaps I'll choose to ignore that for the moment...)
  • Although, in fact, a similar feature could be useful for clipboard operations. We occasionally get complaints about Unix commands pasting incorrectly into PuTTY from Microsoft Word documents; it turns out that Word has "auto-corrected" ordinary quotes to smart ones, hyphens to en dashes, and so on; and when these are pasted into a Unix shell, they tend to come out as dots. So some people would find compatibility mappings on incoming pastes useful, at least.