Re: [Groff] problem with `refer'

[Jon Snader]

On Thu, Jan 31, 2002 at 09:28:55AM +0100, Werner LEMBERG wrote:
> > I don't think compatibility mode alone is correct.  The trouble is
> > that many of us using groff and refer do use some (or even many) of
> > the groff extensions.  To retroactively change refer's behavior from
> > what is has always been in groff (not troff) breaks any documents
> > that depend on the current behavior.  Compatibility mode doesn't
> > help if any of the extensions are used.
> 
> OK.  Let's call that option `-x': neither refer nor groff do use it.
> 

Or perhaps some sort of in-line command.  This might be advantageous
if there are a bunch of compatibility modes as discussed earlier.

> > I feel *very* strongly that if you do this you should have to
> > *explicitly* activate the *new* (not old) behavior.  That way
> > documents written to the current groff still work and those who want
> > the new behavior can get it.  Anything else certainly violates the
> > principle of least surprise, and is sure to annoy those of us
> > depending on the current groff behavior.  For example, I would
> > certainly be annoyed if a new printing of my book escaped to the
> > printers with broken references because refer behavior changed and
> > mangled a reference far away from anything that I changed in the
> > manuscript.
> 
> Are you talking about a real problem or just a hypothetical one?  Have
> you ever made use of this obscure feature?

Well I, for one, use it and I imagine that there are others as well.
This really seems to me like trying to convince the British that their
driving on the left doesn't make sense since it's unlike the ``syntax''
in the rest of the world.  That may be true, but they like it and it
would just cause too much trouble to change it.  The case for making
this change seems to me to be pretty weak--``The syntax is ugly so let's
change it.''  That would be OK if it didn't break existing scripts and
manuscripts, but it does.  What do you say to the user who gets erroneous
results because groff suddenly changed established behavior? How do you
explain to him why this is worthwhile?  Your appeals to syntactical
purity will certainly fall on deaf ears and is very apt to elicit rage
if, as I suggested before, it causes undetected errors to escape into
publication.

The current syntax may be ugly and illogical, but it is established and
users are familiar with it.  I see no reason to change it without
compelling reasons, and so far I haven't heard any.  Those who object
to the current usage can, as you note below, make use of the `[' and
`]' keywords.  Please don't cause trouble for those of us who are actually
using this feature just to make the language into something that seems
new and shinier but really solves no problem.  After all, as Brian
Kernighan remarked, the troff language has always had a rebarbative
input syntax.

> 
> For the sake of orthogonality and to remove an illogical restriction I
> really favour a change of the default behaviour.  Additionally, GNU
> refer already has an alternative, better syntax to specify <pre> and
> <post>, using the `[' and `]' keywords within a `.[ ... .]' block.
> 

As I indicated above, I am against this change, but if we must have it,
then at least make the user choose the change and not have it and the
attendant problems foisted off on him.

As always, Werner, I am grateful for your work and am not merely carping.
I really don't feel this change is beneficial for groff and its users.

Jon Snader