Re: [certi-dev] Bug fixing

[Eric Noulard]

2010/3/3 Yaomin Fu <address@hidden>:
> Hi Eric,
>
> Sorry for late reply. I am off doing other things, and will back to rti
> sometime late.

No problem.
I'm not always responsive myself.

> I'll do the patch once having time.

OK thanks.

> I have tested CERTI with JSAF, using rti-s. Generally it works well, but
> sometimes has abnormal behavior. I will ask our JSAF expert to do a sanity
> test late and will inform you the result.

It would be great.  And we could put some effort to make it work
if it looks reachable and if JSAF team is interested in this too.

> During the "DLC" enabling, what I feel is that to be real DLC, having a
> specification is not sufficient, something at the implementation level need
> to be coordinated. Perhaps that not a specification can do.
>
> For instance, the JSAF uses a RTIFedTime class exposed from rtis. It happens
> CERTI has RTIFedTime class with very similar method signature. Thus, I can
> easily make CERTI DLC enabled (implementing SISO DLC), with minor chance, to
> replace rtis, and make it work with JSAF.
>
> Note that the RTIFedTime class is not part of specification. It just
> implements
> the FedTime interface and provides extra methods to ease the use of FedTime
> object.  If a rti implements the FedTime interface in another way, then
> there
> is no chance to make it work with JSAF, not even possible to be DLC, unless
> add
> to the top of it a wrapping class.

That's perfectly true.
Normally specific FedTime implementation coudl be plugged in any DLC RTI but
won't necessary be able to encode/decode FedTime between different RTI.

> MAK RTI covers this issue by prompting their implementation of the HLA
> interface specification. They have a VR-Link product which allow to 
> dynamically
> connect to any rti if it has implemented the DLC specification. But the users 
> have to
> pay for using their VR-Link to build federates.

This is normal commercial company activities they sell their solutions
and their products,
nothing to be ashamed of.

> I think having a rti code base like yours is a good place to prompt a
> generic implementation.

I do fully agree with that and I would add that I see CERTI more like
a toolbox usable
for building distributed applications than a "product" to be selled.

-- 
Erk
Membre de l'April - « promouvoir et défendre le logiciel libre » -
http://www.april.org