On Mon, 2014-06-16 at 06:43 +1000, NeilBrown wrote: 

tbl I can live with.  Unicode I cannot.
In some contexts Unicode may be ok (non-English words) but not for
line-drawing characters and not for special punctuation.

Well but you know, that tbl(1) won't work for your PDF/HTML rendering
either? At least it didn't when I checked it.

troff should be
able to create those characters itself and you should tell it what you want.

Well I tried that (but I'm not a [n|t|g]roff expert... \[uXXXX]
sequences seem to work fine again for man... but PDF again gives me a
warning: can't find special character `uXXXX'

Actually it seems that at least groff should support plain UTF8
characters and convert them internally to uXXXX representations... (and
as one can see: it does for man)... not sure why it doesn't work with
PDF.

It's not true that "literally everyone" has enough water to drink.
Suggesting they have Unicode is ridiculous.

Well not sure what unicode has to do with water... ;-) ... I've looked
around a bit in Google and found entries with groff/Unicode already
dating back to at least 2006... that's ten years... anyone using very
old or embedded systems which don't support unicode yet, will likely
continue to have a very old version of mdadm manpages (without
unicode)... and those people who really update to a new mdadm (with my
unicode patches) but not to a new system supporting unicode can probably
live with looking at the docs at some newer system.

I mean I don't quite understand why the majority of users should suffer
(and I think using extended characters clearly improves readability)
just for a very small minority... and after all,... that's always some
kinda poor argument of not supporting the new (like Unicode here): "we
don't support it, since there are others as well who don't support it
yet"... kinda self sustaining..

There are certainly web sites that display man pages in HTML, so that is a
minimum requirement.

Well but then again tbl(1) kicks you out of the game. Now of course I
could "hard code / draw" what I do with tbl(1)... and that should work
then with PDF/HTML, when a fixed-width font is used,... but in turn one
looses at man, when the terminal has different sizes.

I think adding proper tbl/unicode support is rather a duty of the groff
guys... and actually there are some mailing list post, where some guy
claims to be working on HTML support for tbl.

The correct approach is not "do what I think is cool" but "do the same sort
of think that all other man pages do to maximize interoperability".

Well again,.. this will keep us in stone age forever,... and I think 
using tbl(1) is a much bigger problem then.
Unicode is THE standard charset nowadays... everything should support
it... if not it's a bug,... especially when looking at
internationalisation it's simply needed... 



Anyway... I guess there's no benefit in discussing over Unicode/tbl
here... :)


I've attached a new set of patches... the second replaces all the
unicode stuff with similar ASCII chars.
I'd suggest to merge both and not just the result of them, so we have
the fancy Unicode stuff in git as well, should we ever decide to upgrade
to post 1991 ;-)


Hope that helps and you can merge them largely as is,... please tell me
whether or not (or whether other tweaks are needed)... so that I can
clean up that branch.

Still have another branch with information you gave me back then, about
how reads and writes are done... I'll come up with that in another mail.

Cheers,
Chris.