[HEADSUP] new portmgr recommendation about adding new ports

Mark Linimon linimon at lonesome.com
Mon Sep 18 23:05:38 PDT 2006


Here is the result of a discussion amongst the portmgr members during the
past few weeks, in response to the large number of new ports added to the
Ports Collection in the last few months.

---------------------------------------------------------------------

Recently we've seen evidence of an increase in the number of ports that
are submitted "just because the software is there", and not out of any
underlying need for the port to exist.

When considering a new port, portmgr would like to ask you to please apply
the following criterion:

  New ports should only be submitted by someone who is actually using the
  software and will continue to use the port to maintain the software.

Adding a port of software that you do not personally use, or do not use
within the FreeBSD ports collection (such as an xpi browser extension that
you manage using firefox and not using the Ports Collection), is not a good
tradeoff between adding functionality and adding complexity.

Each additional port requires resources, both machine (from the package
building system) and people (to inspect the results from the package
building system).  This is especially true when you consider that we, as
a team, are all trying to keep ports buildable on on 4 different branches
and 5 different architectures (with increasing interest in arm and powerpc).

Past experience shows that these ports often quickly fall behind as new
versions are released, and when build breakages occur, they do not get
fixed.  In some other cases, some of our contributors try to keep ports
viable by fixes and updating long after their real usefulness has passed,
and that time could be better spent on the more worthwhile ports.

While no one is suggesting that we go the route some projects have with
some kind of 'gateway' process for approving new ports, at some point the
number of ports will simply be too great for our infrastructure (package
building system, sending PRs via email).  Some common sense should help
to keep us from reaching that point.

As a reminder, the Ports Collection already has 198 ports marked BROKEN*,
and 4291 unmaintained ones.  We really don't need to add to these numbers.

To summarize, we simply can't support all the possible applications out
on the Internet, so we need to use common sense to try to keep it down to
a maintainable number.

mcl

(*: on i386-6; the numbers are higher on the other architectures, and -CURRENT)


More information about the freebsd-ports mailing list