Re: dselect complaints

["William S. Gribble" <grib@cs.utexas.edu>]

Ian Jackson wrote:
> I'm not interested in hearing any more complaints or even extensive
> suggestions for improvement, unless the person complaining is
> volunteering to do the work on a new interface.

If you don't want feedback about the tools, might I suggest that you
give up their maintenance to someone who does?  Dselect as it exists
is nothing more or less than a working prototype of the tool it needs
to be.  

Dselect has come a very long way since the days of the perl dpkg, and
the work you have done on it is valuable and admirable.  However, your
statement suggests that you regard dselect as being a finished product
which needs only some debugging.  I don't agree.

I realize that it is too late for a rewrite of dselect before the
release of 1.1.  However, let's just get it out in the open: as it
stands, dselect is a liability with respect to general acceptance of
debian.  It is a nightmare to use, has an interface that only its
programmer could love, and gives the appearance that at every moment
you use it your entire system is in jeopardy.  The subtext in Bruce's
recent public announcement about the relationship between Debian and
the FSF was that Debian is comparable to or better than Caldera/Red
Hat.  New users are going to look at dselect and immediately decide
that Bruce must be smoking crack.

What needs to happen, in my opinion:

1. Make the release of debian-1.1 public, when outstanding bugs are fixed.
   Emphasize alternatives to dselect in the documentation.  Make it
   clear that dselect is still a working prototype rather than a finished  
   product.  Provide script(s) to handle several common initial-installation 
   scenarios, with instructions about how to use dpkg to install more or 
   remove unwanted packages. 
2. Open a period of discussion to draft a new specification for dselect, 
   using the lessons learned from the current version as a starting point. 
3. Rewrite dselect to meet the new specification. 

Ian, please don't tie your ego up with the dselect code.  You will
only give yourself ulcers.  Listen to what people say, incorporate
their suggestions, and Debian will be better for it.  Dselect is
almost 100 percent about user interface, and if you don't listen to
your users you aren't doing your job as a programmer.  You have done
good work so far, just keep going.

Bill Gribble