I realize that this discussion has slipped towards a Maya vs. XSI battle royale, but I believe it started out as a recomendation to a company. Maya is an awesome package, and some really amazing things came out of it, and no doubt will conitinue to even if it dies tomorrow. The thing is, as a small business, I still wouldn't buy Maya, even for $2,000. Because the workflow in XSI is so much faster.
Markman points out that MR isn't the default film render, and that there are others coming up fast...but that's not really relevant to the issue of workflow. I've seen great stuff out of the Maya renderer, but it takes too long to tweak. It is the integration of the render tree with MR that makes it such a powerful tool for production. Sure you can make a gorgeous image with Entropy, and it might even shave a few seconds off of the render time, but the workflow won't be as fast.
Frankly, I'm happy if nobody wants to look at the workflow issue, because it mens easier competition for those already using XSI. I am confident that in the non-film industries a darwinian natural selection process would result in most video post houses using XSI because they can turn out better work _faster_.
But Matt's scenario of "trickle-down Mayanomics" scares the piss out of me, because I could indeed see such a thing happening. Hopefully with better NURBS tools (2.5, please?) and a renderman pipe more film companies will use it for te same reason, quicker job turn around. The excellent scripting options should encourage integration of XSI into there pipelines. Here's to hoping for a bright XSI future.
Scott