Trivial +server variants considered harmful

[Ryan Schmidt]

On Oct 26, 2008, at 22:49, William Siegrist wrote:

> On Oct 26, 2008, at 8:21 PM, Joshua Root wrote:
>
>> There seem to be some ports (e.g. ntop, mysql5) with a +server  
>> variant
>> that does nothing but create a startupitem. It seems to me that it  
>> would
>> be much better to remove the variant and just always create the
>> startupitem, since the cost of creating a startupitem is negligible,
>> while the cost of having to recompile the port if you forget to  
>> specify
>> +server is considerable.
>
> Agreed. And if the server functionality is expensive to build, I  
> prefer the postgresql model of having *-server ports that can be  
> built separately when possible.

It's possible to disable building the server parts by adding -- 
without-server. The mysql5-devel port does this; it was requested in  
this ticket:

http://trac.macports.org/ticket/14146

However, that's incompatible with the desire to have a separate  
mysql5-server port as requested in this ticket:

http://trac.macports.org/ticket/12313

The server variant was added to mysql5 in this revision:

http://trac.macports.org/changeset/13929

I can't say I know why it was added as a variant and not by default.  
I'm not opposed to changing it.

There are those who complain when ports cannot be installed by non- 
root and then we end up adding +no_startupitem variants, like the  
apache2 port has. Comments on this?