On 15/11/21 09:42, Rob LA LAU wrote:
Hi,
On 14/11/2021 20:49, Guido Falsi wrote:
You talk about "adding a periodic script". That is not even a real
modification to the upstream software IMHO. Just adding some glue code
for FreeBSD. If the script does what it advertises, and has no
malicious intent I see nothing wrong with it. If it is broken fixing
it is the logical thing to do.
Imagine a daemon, an rc script and a periodic script.
And imagine that we decided to separate all 3 into separate ports.
When fitting these ports into the ports tree, we would make the daemon
depend on the rc script, because the daemon needs the rc script to start
and stop (and starting and stopping could be considered core
functionality).
But we would make the periodic script depend on the daemon, because the
daemon runs fine without the periodic script, but the periodic script is
useless without the daemon.
To me this would mean that the daemon and the rc script should be
published together as a single port, making the rc script an acceptable
FreeBSD-specific addition. (Which does not mean that the addition should
not be reported to upstream; the goal should also be to have the rc
script included in the upstream project, just like the systemd unit that
is already included.)
The periodic script, however, should be considered new functionality,
and published in a separate port.
And to contradict the above, one could now plead that external libraries
that the daemon depends on should also be packaged together with the
daemon, but the difference obviously is that no other package will ever
depend on the daemon's rc script, while those libraries have been
created to allow multiple applications to use them.
And there will always be exceptions. Things are not always black and
white, and they shouldn't be.
And obviously, I'm the outsider here. I'm not trying to tell anyone what
to do and how to do it. I'm just making sure that this subject has not
been overlooked.
I don't follow you. The example is too abstract for any useful purpose.
You intentionally mix the concept of "depending on" with the concept of
"packaging together".
Even following your example all the practices you describe could be
acceptable and at the same time debatable. The variety of software is
too broad to create a strict rule set. Such matters need to be handled
on a case by case basis anyway
We do have some principles to guide us, which have been stated to you.
One of these is classic POLA, for example. This gives us an idea. You
have to consider how the original software is packaged upstream and
usually packaged for other OSes. Considerable differences from the rest
of the world best practices would be "astonishing".
Also you have to consider upstream documentation. If that suggests doing
things in a certain way and the port did things in a different way that
would be astonishing too.
In the end you also seem to forget or intentionally ignore that FreeBSD
is also a community of developers and users. You do things the best you
think, but it can happen that users or other developer will tell you you
are wrong, maybe suggesting better solutions. Ideally this is done in a
polite and constructive way; unluckily it sometimes gets aggressive and
unpleasant, and this is definitely something we should avoid.
You can then modify your work and make it better based on this
feedback., No technical choice is set in stone.
Being a son of two lawyers, and having a lot of friends who are
lawyers, and also clients who are law firms, my informed (by having
had a lot of arguments with all these people about this very subject)
opinion is that there is no such thing in the world as a "clear and
simple" set of rules.
Nobody is trying to write a legal document here.
The goal is just to assure that the ports in the ports collection still
function as intended by the upstream developers, preferably with no
functionality removed, and definitely with no functionality added.
You look too worried by the "functionality added" part. If the added
functionality is only a script needed for integration with the system
it's not added, but something strictly needed by the porting effort.
For example I maintain some ports (asterisk, ntopng) which are run as
daemons. Upstream does provide startup scripts for linux, but none for
FreeBSD. The ports contain startup script I made because they are needed
for proper usage of the port.
(and some bugs have happened in those scripts, I have fixed those to the
best of my ability)
Users are not force to use the startup script I (or upstream for that
matter) provides. They are disabled by default. Nothing stops the users
from writing their own rc scripts, or using a different method of
starting them (cron @reboot or something like supervisor for example).
Those scripts are "added" but not mandatory for anyone, they don't
detract anything from the upstream software.
I don't know which port periodic script you are talking about, but I
guess it's not enabled by default, and in the unfortunate case the
maintainer went out of his way to make it such, it should be possible to
disable it, and then you can script your own or use something else for
that part of the port.
This for startup and rc scripts...other more convoluted cases require
case by case analysis. A more general reply is impossible.
--
Guido Falsi <madpi...@freebsd.org>
