Re: New module decisions for 2.22

[Alp Toker <alp atoker com>]

Vincent Untz wrote:
> cc-ing release team to get input from others.
>
> Le samedi 12 janvier 2008, à 03:38 +0000, Alp Toker a écrit :
>   
> > Vincent Untz wrote:
> >     
> > > Hi,
> > >
> > > Le vendredi 11 janvier 2008, à 23:29 +0000, Alp Toker a écrit :
> > >   
> > >       
> > > > Alp Toker wrote:
> > > >     
> > > >         
> > > > > >  + ndesk-dbus, ndesk-dbus-glib (external dependency)
> > > > > >   - good from a security point of view
> > > > > >   - need to be in a mono-specific section of the external dependencies
> > > > > >     (because of the special rules about depending on mono)
> > > > > >   => accept
> > > > > >         
> > > > > >             
> > > > > Hi Vincent,
> > > > >
> > > > > I'm not too comfortable having this in a Mono-specific section. Managed 
> > > > > D-Bus isn't a Mono project for good reasons and I've taken care not to 
> > > > > depend at all on Mono or other non-standardised platform libraries. This 
> > > > > is pure ISO-specified C# code as much any C module in the platform -- it 
> > > > > can be considered a Free replacement for non-standardised .NET Remoting 
> > > > > features.
> > > > >
> > > > > The IPC system based on two freely available freedesktop.org 
> > > > > specifications (XDG[0] and D-Bus[1]), both of which are widely 
> > > > > implemented and deployed without issues.
> > > > >
> > > > >  [0] http://standards.freedesktop.org/basedir-spec/basedir-spec-0.6.html
> > > > >  [1] http://dbus.freedesktop.org/doc/dbus-specification.html
> > > > >
> > > > >       
> > > > >           
> > > > Sorry, I forgot to explain UnixMonoTransport.cs which may have thrown you 
> > > > off. It's still in the git repository but isn't shipped in tarballs. It's 
> > > > been rewritten to use standard Unix interfaces in the file 
> > > > UnixNativeTransport.cs specifically so managed D-Bus can become a fully 
> > > > blessed GNOME dependency.
> > > >
> > > > You can find out more about the clean room approach to development in my 
> > > > slides from GUADEC:
> > > >  http://www.atoker.com/dbus/managed-dbus-guadec07.pdf
> > > >
> > > > An update on-list clarifying that it's an ordinary blessed dependency 
> > > > without special status would be appreciated.
> > > >     
> > > >         
> > > We should probably talk about this with the whole release team, but
> > > here's the rationale:
> > >
> > >  + AFAIK, ndesk-dbus needs mono to be used
> > >       
> > I did mention, it doesn't need Mono so we can avoid that whole debate. I 
> > put care into making sure managed D-Bus works on multiple CLR 
> > implementations throughout the development cycle.
> >
> > It's been known to run under DotGNU (somewhat slowly) and I know there's at 
> > least one educational project using it with Rotor (they're using D-Bus to 
> > remotely control their robots; fascinating project using a GTK+ frontend 
> > and managed D-Bus -- all done without Mono).
> >     
>
> I quickly looked at the configure.ac, and it still seems to require
> mono, via the Mono.Posix assembly, eg.
>
> But I see your point, and that's an interesting one: DotGNU wasn't
> considered when creating the mono rule.
>   

Mono.Posix is just a thin wrapper around Unix syscalls and doesn't 
depend on Mono or any non-standardised CLR extensions. If having "Mono" 
in the name is still a problem for the release team, I can look into 
copy-and-pasting the 200 lines or so I need directly into managed D-Bus.

Managed D-Bus is written in ISO/IEC 23270 C#, just like we write other 
parts of GNOME in ISO/IEC 9899 C. It's built up on freedesktop.org 
standards, some of which I've helped to author -- this is pretty much as 
free as free software gets.