Re: info -> html

[Patrice Dumas]

On Wed, Mar 21, 2012 at 05:00:49PM -0700, Karl Berry wrote:
> Chong (and Patrice and Sergey and Stefano),
> 
> Sure, sounds like a reasonable starting point to me.

Taking a step back, maybe it would be nice to think about the output
languages we want for the different use cases.


We have texinfo.tex for pdf.  Adding a LaTeX output would be a plus as
it allows to do things TeX cannot and since it is very similar with 
Texinfo the conversion should be rather simple, but it is not high 
priority as the texinfo.tex output is already very good.


Then we have html for online browsing, and, I guess, for small graphical
connected terminals, like mobile phones.


We also need something for browsing documentation locally on desktop and
on the console.  Both have different requirements.

We could use docbook for the desktop and use the infrastructure of 
gnome/kde for documentation.  I cannot speak for the emacs people, 
but I guess there are good reasons to have emacs be able to display 
docbook, be it generated from Texinfo or not, especially on the 
desktop, maybe that's what direction is pushed on the emacs side.
Then we may want to think about ways to make Texinfo more compatible 
with DocBook, maybe first with more warnings when converting to 
DocBook when the output is not correct DocBook.

We could also use HTML.


Then we have the console.  Info is a good format for the console, in my
opinion because of the following characteristics:

* it is pre-formatted.  The display program do not have to reformat
  text.
* it is 'indexed' by the table of nodes, so that there is no need to
  find the cross references destinations, only the starting points.

The consequence is that the standalone info reader may be quite simple
which is good when there is a need of a lightweight reader that can 
compete with man pages for admins for instance.

It has also some disadvantages.  First, in the default case it is
adapted to a given text width (which in in contrast with xml output 
formats that are not specific of a width).  Second it is a simple 
text format which renders coloring the output, for example, not easy 
(although maybe we could use some terminal escape sequences at 
generation time).  There is very light markup which is both an
advantage since it allows the reader to be very simple, and a 
limitation, especially since punctuation and * are both markup and 
text, and here I refer to the recent threads I launched about the
issues with punctuation in nodes.


We can use HTML or docbook for the console.  (We can also use Texinfo
XML or any other intermediary output that maps better to the underlying
Texinfo, but with the issue that we'd be doing things by ourselves.)
But then the reader needs to be more complicated because the 2 items 
I list above are not available anymore.  For this more complicated
reader, we can use emacs, or an existing html browser, but with the 
caveat that they may not be relevant for a minimal install and may not
be appealing for people liking minimal setups (such as myself ;-).


In any case, I think that we should discuss the requirements of an
output format separately for the console and the desktop.  On the
desktop we should certainly try to follow on what is happening on
freedesktop, gnome and kde, even if we know that it changes a lot
and will involve some work to keep up with the pace of change here.
On the console we have to decide whether we keep something specific like
the Info output or we adapt to the desktop formats.


Also maybe we should move that conversation to bug-texinfo as soon as
possible?

-- 
Pat