Re: Shipping Vera with 2.4

[Havoc Pennington <hp redhat com>]

On Thu, Feb 27, 2003 at 06:20:08PM +0000, Gustavo J. A. M.  Carneiro wrote: 
>   [Disclaimer: this is just my mostly-uninformed opinion.]

You should read the message-bus-list/xdg-list archives, and probably
also if we want to discuss this the thread should start with a
detailed analysis of the big picture. This piecemeal discussion isn't
going to be very informative or get us to a conclusion.

>   If you have IDL and stubs/skels, then is it much different from
>   CORBA?

In some ways it is, in others it isn't. 
 
>   Is ORBit slow? If it is:
> 	1. I'm sure it can be optimized.
> 	2. It doesn't matter *that* much. Even if you made ORBit 10x faster,
> would it make GNOME faster? I doubt anyone would notice the
> difference.

No, in fact I expect D-BUS to be slower than ORBit. D-BUS is currently
about 2x slower than ORBit in the peer-to-peer case, and introducing a
message bus in the middle gives you three processes instead of two so
is slower than that. Optimization will probably bring D-BUS into a
similar performance range as ORBit but it's not going to be faster.

The only performance win for D-BUS may be that it has more of a focus
on avoiding round trips.
 
>   Does unification at this level bring any end-user advantage?

It certainly does. Of the Linux desktop usability problems right now,
quite a few of them are caused by lack of standard ways of doing
various things, such that all apps are living in their own little
world.

Take "recent documents" feature of GNOME 2.2 - it completely sucks
because of the big-three end user apps Mozilla, OpenOffice, and
Evolution, exactly zero of them use the "recent documents" feature.
There's a file-format-based spec for recent files now, and I guess
people are adding OpenOffice support, but it would have been much
easier with a D-BUS style system.

Havoc