Subject: Re: time spent per package
To:, Lubomir Sedlacik <>
From: None <>
List: tech-pkg
Date: 03/11/2003 12:46:15
Thanks for the responses.  The large range confirms my own experience,
but I wasn't really sure if people who make more packages than I
generally are quicker.

Lubomir Sedlacik writes:
 > maybe Brook could tell us where is he aiming with all of this?  because,
 > honestly, it looks at least a bit strange (should we sleep less to make
 > more packages a day?  or at least skip a dinner? :)).

Clearly people can only make what packages they have time for, and
clearly individuals need to make decisions about how they spend their
own time.

However, I am trying to raise some resources that might either help
the process or provide monetary incentives for anyone interested in
certain broad classes of packages (primarily scientific applications
of which I have a number of specific examples).  This seems to be a
particularly difficult area, probably for the reason Hubert points
out: not only does it require packaging skills, but it also requires s
knowledge of the science involved in order to debug the packages.  I
am involved with a number of scientists who can provide some of the
latter expertise if I can catalyze some input from the packaging crew
for the former.

My only hopes with this were that the people involved with packaging
might be compensated a bit for their efforts and that the NetBSD
package system would benefit as a result.  If this is "out of bounds"
please let me know and I will pursue other avenues.

In order to quantify the resources that might be required to
accomplish this, I needed a bit of information on how long each
package might take someone with experience making packages.  Hence my
first question.  Thanks alot for the answers; they are very helpful.

Supposing I can identify the needed resources, the second question is,
what might be the best mechanism for directing them to people with
interest in expanding the set of scientific application packages?
Presumably, the process would involve some iterations between
creating/updating a package by someone with that expertise and testing
it by a scientist, with feedback going in both directions.  Perhaps
there are other approaches; I'm open to suggestions.

In any case, please make suggestions concerning how best to initiate
collaborations or interactions that would serve to increase the set of
scientific application packages available.  Some relevant questions I
can think of include:

- Is there any interest in packaging scientific applications?
- Should the packager(s) be compensated in any way?
- If so, how should the compensation be determined or on what should
  it be based?
- How should the nature of each package be defined (e.g., some are
  self-contained, whereas others should be subdivided into several)?
- How can specific deliverables be defined?
- Does every single package need to be negoitated individually, or is
  it possible to apply "average" terms to many?
- Is this even a useful or desirable approach?
- What would be an ideal procedure?

I hope this will serve to stimulate a useful discussion that will
identify a foundation for enhancing pkgsrc.  Thanks for your
patience and thoughtful responses.