[Spread-users] Re: [Software-bus] What are *you* looking for in a software bus?

[Ned Konz]

On Tuesday 06 August 2002 10:31 am, Jason W. May wrote:
> In the environment you described for Microship, I wonder if this
> assumption still
> applies.  If you have to deal with changes in latency between
> nodes, node availability,
> and transmission speed, what impact does that have on the
> requirements for a message bus?

I'm not sure that it impacts the bus itself at all. For instance, 
using something like Spread, we'd eventually hear that a group had 
lost connectivity between one of its members and the local daemon. I 
*think* that Spread can handle having the network partitioned and 
then re-attached. Assuming that really important messages are held 
until the network connectivity is restored, the bus should work 
transparently.

However, I can see some value-added features that could be provided by 
such a system:

* message priority classes or bandwidth preferences. Don't bother 
sending unless the priority is high enough or the bandwidth is high 
enough to all the group members.

* notification on significant changes to topology or speed for the 
membership of given group.

* queries for effective bandwidth to members of a group

-- 
Ned Konz
http://bike-nomad.com
GPG key ID: BEEA7EFE