Adding software to base

[Thor Lancelot Simon <tls%panix.com@localhost>]

It seems to me there are several common-sense criteria for adding
software to base.  None are entirely objective.

1) There should be a clear benefit to having the new program or
   library in the system for a substantial number of NetBSD users.

2) The new program or library should provide functionality that is not
   already provided by some other program or library in the system.

3) The new program or library should be small, even when considered along
   with any dependencies it brings to the system.

By these criteria, from my point of view, if we were considering adding
each today, tmux would clearly come in, while the huge gonkulating beasts
that are Taylor UUCP or sendmail would be out -- the former on criteria
#1 and #3, the latter on criteria #2 and #3.

In any case, it is a matter of proper balance.  Keeping the base system
compact and simple is a very important goal but *it is not the only goal*.
Adding programs used by a large number of NetBSD users to the base system
is a very important goal but *it is not the only goal*.  Avoiding
duplicated or confusing functionality is a very important goal but... you
get the idea.

I may not like the choices that are made in this area sometimes (e.g. tarting
up 'ls' with every poorly-considered option any Linux user ever added to it)
and others may not like the choices *I* make (for example, importing a new
compression library on the grounds that it is _potentially_ useful but that
the barrier to entry should be quite low because it is extremely small).  What
is silly is to insist that only one criterion matters, or that it is not
possible to reach a good compromise.

I also (obviously) think that the right answer for the system as a whole
does sometimes involve taking large, lightly used pieces of code _out_.

Thor