> Branko ??ibej wrote:
> >Looking at this from the point of view of the typesetting world,
> >there's no reason to introduce long lines in the source. Within
> >running (roman) text, the optimal line length for easy reading is
> >about 68 glyphs; anything wider makes it harder for the eye to skip
> >to the next line.
> Yes - 68 plus or minus 20 or so - it's not an exact science. Thus,
> this isn't a compelling argument in itself for reducing from 79 to 68
> or whatever, but it does provide a "feel-good factor" if we have
> another reason to reduce the width, and we do.
Right, 79 isn't *that* much greater than 68. And in a lot of option
help, lines are effectively a lot shorter anyway, due to the 2-column
> Also, in a shell window, in contrast to printed matter, we need to
> make reasonably efficient use of the horizontal space available, so
> as not to use up unreasonably much vertical space, so we shouldn't
> use a width much less than around 60 to 70 columns.
Efficient use of vertical space is important in printed matter too.
The difference is that in printed matter, when your text gets too wide,
you split it into two or more columns. On a CRT, with only 80 columns
to work with, that's not really possible, except with things like
option help where the 2-column layout is already natural.
> We could now shorten the displayed lines so as to bring the source
> back to 79 columns.
If I had a vote on these matters I'd strongly suggest that optimally
formatted output is much more important than optimally formatted source
code - and that allowing long lines in the source (as the patch in
question did) is the least painful way to achieve this.
Also, I agree that citing 68 glyphs as optimal is misleading, in light
of both vertical efficiency and existing indentation (due to 2-column
help text, or other reasons something might be indented a few spaces).
Remember, the indentation doesn't count toward a line's readable
In cases where text can be reformatted to make lines shorter than (say)
70 columns, *without adding any lines*, I'd be in favor of that too.
Received on Tue Nov 29 16:55:30 2005