Re: [certi-dev] RTI1516 - NameReservation set to be distributed?

[Eric Noulard]

2010/4/29 Ijperen, Jeroen van <address@hidden>:
> Hello,
>
> As part of my work on the RTI1516 section for CERTI, I am currently working 
> on implementing the 'Object Name Reservation' service. As you may know this 
> is a service that allows Federates to request a reservation for names for 
> objects, after which it receives a callback indicating whether or not the 
> reservation succeeded.
>
> I have figured out that the set of name reservations should be kept in the 
> RootObject of the RTIG. My question now is if this set should also be 
> mirrored/distributed to the connected RTIA's?
>
> While this is perfectly possible by broadcasting the Name Reservations, I 
> doubt it's use, since all Name Reservations and Object Registrations are 
> handled solely by the RTIG.
>
> Any thoughts?

I think you are right about the RTIG handling, so I don't think there
is a need for duplication
of the information inside RTIA.

I see two uses cases in favor of a RTIA duplication

1) we want to be able to send fast "Error" answers
for already reserved names **without** asking to RTIG.
i.e. if RTIA already know that the requested name has already been
reserved there is no
need to check RTIG, RTIA may be able to answer on his own.

2) we want to support the run of federation without RTIG

The 2) is an interesting goal which may be worth studying in the future.
The idea would be to replace the "centralized RTIG" with a
"fully distributed RTIG".
This may be possible for many HLA services but the time service.

I think 1) is not currently worth the effort and 2) is not a *current* goal
so I would advise to keep "name reservations" design centralized in RTIG.

This is my own thought, may be others have different idea concerning
this service?
-- 
Erk
Membre de l'April - « promouvoir et défendre le logiciel libre » -
http://www.april.org