a cabal for D ?

Bruno Medeiros brunodomedeiros+spam at com.gmail
Fri Apr 1 11:27:55 PDT 2011


On 18/03/2011 18:42, Russel Winder wrote:
> On Fri, 2011-03-18 at 10:41 -0400, Jesse Phillips wrote:
> [ . . . ]
>> Otherwise I would be for using the native packaging system, but it
>> isn't easy for the contributor, the maintainer of the server, or the
>> one building a system that works with all of them.
>
> The problem is that the system administrator wants to use just the OS
> packaging system to make things easy for them to maintain the system.
> The user wants to be able to install stuff easily and may not have
> sufficient permissions to actually use the OS packaging system.  Sys
> admins don't want to have to learn N packages to deal with N languages,
> the use probably only works in a couple of languages and so doesn't care
> if they have to learn two different packagin systems so as to get stuff
> done.
>

I agree sys admins should not have to learn language specific package 
managers just because the user wants to use one. Rather what should 
happen is that language PMs should be able to be "installed" on any 
user-specified location, and installed multiple times. Just like a 
compiler should. They should not be designed to work only as an OS 
"singleton", so to speak.

> This is not an easy issue.  I just find the knee-jerk reaction of "we
> have this new language therefore we must have a brand new (build system|
> packaging system|shell|implementation of every comms protocol|new user
> interface library|operating system)" leads to too many distractions from
> getting stuff done using the good tools that are already available.
>

What I think makes sense for a D package manager is to integrate with 
existing *build* systems (make, Cmake, Ant, Maven, SCons, etc., etc., 
whatever might be appropriate). But definitely not with OS package managers.

-- 
Bruno Medeiros - Software Engineer


More information about the Digitalmars-d mailing list