Fellow Simulation Folk, we stand at a cross-roads:
It is easy to yearn for the simpler times, and even to enjoy a visit to them. I own a 1985 Mustang GT, the last American muscle car with a carburetor. I am currently bringing it out of 4 years of storage. However, It is clear that my expectations may have changed in the interim! So is the state of SISO!" However, I am being careful to keep it what it is. It is (was) the coolest car that a Midwestern 25 year old could own in 1985. The temptation to dump $15K into it to make it incrementally more capable is strong. However, this would dishonor the memory of the technological achievement that it represents. I would rather keep it as it is, and be passed by serious drivers in late model Camaros who are in a rush to enjoy their youth. If I want to go fast, I have a 1999 BMW motorcycle that will crush that Camaro, at half the cost, and twice the fun!
The (traditional SISO) simulation community is in a similar situation. DIS is a beautiful and beloved 1985 Mustang GT. It represents raw UDP, distributed real-time training simulation interoperation simplicity at its best. I still love it! However, like my ’85 GT, it is better remembered and visited than used for a workhorse vehicle. I fear that we are taking the opportunity of the HLA and "dumbing-it-down" for nothing more than emotional comfort. As an example of this, I have watched the saga of the "short-term" effort on the Real-Time Platform Reference (RPR) Federation Object Model (FOM), RPR-FOM, with interest. I thought that the attempt to capture hard-won DIS integration experience, by and for the real-time distributed training community, was, and is admirable. I also think that struggling to maintain a sense of community (federation?), and develop a community-wide approach to HLA application are the right things to do. However, I have come to view "The RPR Approach" as too nostalgic. I keep hearing someone playing Sinatra’s "I did it MY way" in my ’85 GT. It does not fit the vehicle, or the current times.
Being a fair-minded individual, I will not let the HLA and RTI out of this discussion without an honest look either. The HLA does achieve the goals of giving almost every distributed simulation interface a hook/function to everyone who wants it. For those of you spending significant time on Mars recently, the HLA is a stunning achievement that I am proud to be associated with in a small way. However, the HLA is more like the power-plant in a new Cadillac than it is the old Mustang. No more room for shade-tree mechanics here. Now it is all tightly integrated electronics, and throughput-optimized voodoo. It used to be that if you wanted to hop up your car you tacked on a pair of headers, and headed for the dragstrip. Today if you try this you mess up the dynamic flow tuning of the integrated-computer-monitored-intake/exhaust track, and you actually go slower, if at all. Little surprise that the RTI is like this. Just a consequence of the times.
The HLA and associated RTI implementation is a carefully and tightly integrated application which can do nearly everything, in certain combinations, if configured correctly, assuming you avoid the limitations and contradictions of trying to graft together disparate simulation approaches. I am also a bicycle rider, and am doing a 100 mile bicycle ride tomorrow (May 15th). When I went to buy a portable, universal bicycle tool kit, I had a choice of kits that were essentially one thing that did everything, and kits that offered several tools which fit cleverly together. I chose the latter because no matter how many tools I have on a single magical toolkit, I can only effectively use one at a time. I guess I am saying that I think that the RTI, and possibly the HLA, is a little too integrated. Maybe it is just a little too good! I think that many people would be happier with a "Lego" set than a "Transformer" toy. It would just leave a little more to the imagination.
So, where does this leave us? Love it, or hate it, we have a simulation integration architecture, the HLA. A long time ago I learned that there is no such thing as bad technology, just bad applications of limited technology. I have also learned that all technology is limited. DIS was simple and effective, but severely limited. Its Achilles heel was that it was optimized for one simulation community. The HLA and RTI are powerful and complex, but monolithic. However, these facts are no reason to hide in a safe corner and pretend we can re-create simpler times.
Let us stop dancing around dying fires and look to the future. We need to empower (simulation) communities to thrive while ensuring inter-community commerce (reuse). This was my intent in choosing the style of the Reference FOM Study Group. We were not going to come back from the SG with a conclusion that the HLA was bad (because it is not) and we were not going to come back with a one-size-fits-all standard dictated by the largest/most-vocal/best-represented community, because one size does not fit all. (Sorry Warren, we just do not agree here!)
Let me explain a simple, yet widely misunderstood fact about Reference FOMs. The intent of the RFOM study was to make sure that if a community asked for rope, they got it. This is analogous to freedom of speech. I may not agree with what you say… What you do with it is your responsibility! With the requested rope, a particular community could tie themselves up, tie themselves to an anchor, or build a bridge, as is their right, as Americans, to "Life, Liberty, and the Pursuit of (simulation) Happiness"…
My engineering and modeling roots may start to show here.
A Monolith for a Monolith. Perhaps we have reached a local maxima in distributed simulation integration optimization. We have a monolithic RTI and one (several?) communities have developed monolithic content descriptions (FOMs) to optimize their use of it. Back before I bought my Mustang I learned that the brightest people that I knew considered at least two options before deciding, and the very brightest always considered 3! I believe that there are other local opmtima. I know that less monolithic approaches have been suggested for both RFOMs and RTIs. I would rather that we start looking to the future, rather than optimizing the present or recreating the past.
Prepare to go to metaphor warp…
Like everyone, I see the crystalline purity of monolithic solutions. However, with the 30-something realization that I am mortal, came the understanding that I could not do it all myself. Like my old GT, there is no perfect anything, just amazingly cool things in their day. I used to worry about real-time performance, and bandwidth. However, I have given in to Moore’s law. I now try to architect systems which will perform at about 0.25 real-time on 0.25 of the available bandwidth. The underlying technology will bail me out! Our technology is not like stationary target shooting, this is more like skeet shooting where you swing-through the target to hit it. Excuse the rambling of an aging modeler who is immersed in the component revolution. But I feel obliged to say that we should enjoy the memories of the past. I cannot wait to take the T-tops off my old Mustang GT and run it down a twisty road with a Billy Idol tape playing. However, we need to successfully and adaptively use the things of the present. I really feel the need to build a lunchbox example of Base Object Models (BOM) based on the RPR-FOM. We should use our insight, education, imagination, and most importantly our optimism to build a better future. Maybe I should look at a radical 100lb-weight-reduction/20hp-increase project for my BMW motorcycle, there is a lot of titanium and carbon fiber components floating around out there!
This monograph is very much my opinion, but honest, heart-felt, and meant to benefit the community!
