Subject: Re: RFC: recommended dependencies (diffs attached)
To: Rene Hexel <rh@netbsd.org>
From: Adam C. Migus <adam@migus.org>
List: tech-pkg
Date: 01/08/2004 12:34:36
On Thursday 08 January 2004 5:49 pm, Rene Hexel wrote:
> On 08/01/2004, at 7:26 AM, Adam C. Migus wrote:
> > How about basically doing what Rene said but changing the semantics
> > a little bit so that the BUILDLINK_RECOMMENDED become the
> > BUILDLINK_DEPENDS and the old BUILDLINK_DEPENDS for cases that
> > caused this discussion could go to BUILDLINK_DEPRECATED (of course
> > alternate naming suggestions welcome).  Thus changing
> > IGNORE_RECOMMENDED BUILD_WITH_DEPRECATED (naming suggestions
> > again).
>
>    Except that this would require changing every single existing
> buildlink file, whereas leaving the old name for BUILDLINK_DEPENDS
> does not require any changes unless a package gets updated and
> recommendations should be set.

No, it wouldn't.  I'm guess my explanation either doesn't doesn't 
adequately explain my idea or you simply don't understand.  Either way 
it would only affect the packages in question not all of them.

>
> > The ideas being that 'the norm' stays the same, the exceptional
> > case becomes apparent and in time the _DEPRECATED can be removed.
>
>    Except that this is a different implementation.  This requires
> a semantic change to buildlink.
>
>    Except for maybe the names, I'd prefer the simplicity of my
> original proposal ...
>
> > In addition (if not already the case) it would be nice if the
> > package data contained the information that it was build with
> > _DEPRECATED.
>
>    Yes, that would be nice, too.
>
>    My suggestion would be to implement the simple approach
> first (as it works without any changes to the package tools),
> and in a second step to add to the binary packages flags
> such as INCOMPATIBLE and DEPRECATED (they are independent
> to a degree and as such can easily be added later on if
> we want to go further with this).
>
>    Cheers
>        ,
>     Rene

Yeah, that sounds about right.

-- 
Adam C. Migus - http://people.migus.org/~amigus/