D-Bus Broker Announced As A New, High Performance Message Bus

[F.Ultra]

You know, if you guys are taking the time to get a new general purpose message bus in the kernel, then you really should design a new one. I mean dbus is not modern. It sucks ass. It's broken by design. It's slow as fuck by design. The saddest part about it is that the team who design dbus in the first place almost had to do it on purpose knowingly.

That ship sailed many moons ago. There are too much software tied to D-BUS now for anything new to make a dent.

[F.Ultra]

D-Bus cannot possibly make use of high performance capabilities because of inherent design limitations.
From a client point of view, it uses a single connection with no internal locking or queuing. Therefore, the only robust way to communicate over DBus by running dispatcher in a single thread. Message rate will suffer regardless of how fast the server is.

Getting rid of server hangs/crashes would be nice though.

No internal locking is good since it means that single thread applications does not have to pay for the unnecessary locking overhead and it's dead simple for the user to add their own locking on top if they know that they are going to use it multithreaded. Queuing would not solve anything since that would equate "running dispatcher in a single thread".

[duby229]

That ship sailed many moons ago. There are too much software tied to D-BUS now for anything new to make a dent.

I don't think backwards compatibility is a good excuse to keep something retarded. Einstein would call that insanity.

[starshipeleven]

I don't think backwards compatibility is a good excuse to keep something retarded. Einstein would call that insanity.

Adding "Einstein" to your claims doesn't make them smarter. If something has reached a critical mass then it's very hard to replace it.

[duby229]

Adding "Einstein" to your claims doesn't make them smarter. If something has reached a critical mass then it's very hard to replace it.

Ok,, so lets all of us entrench it more..... Yep, definitely insanity.

[bonbon]

Sorry if my question sounds a bit n00b, but what is the advantage this solution brings compared to dbus being totally implemented in userspace?

Isn't it also a little... Adventurous to go that route now, given that Bus1 is still far from being finished, and even further from being accepted into the mainline Linux kernel?

[starshipeleven]

Ok,, so lets all of us entrench it more..... Yep, definitely insanity.

The fact that you still can't grasp that when a window of opportunity closes you're pretty much shafted is amusing.

Just look at MS (or Intel) trying to get into the mobile market way too late, and how well it went even if their offering was on par or even better than the competition.

[bonbon]

Apologises if my question sounds stupid, but what exactly are the advantages of implementing DBus on top of Bus1?

[duby229]

The fact that you still can't grasp that when a window of opportunity closes you're pretty much shafted is amusing.

Just look at MS (or Intel) trying to get into the mobile market way too late, and how well it went even if their offering was on par or even better than the competition.

That assumes on par competition. MS never came close.

[F.Ultra]

I don't think backwards compatibility is a good excuse to keep something retarded. Einstein would call that insanity.

So let's replace say x86 with a new modern architecture, call it IA-64 and see how it will take over the world!