The Pod: software issues
Since the porting series will be about application software to be run on this thing, the question here really comes down to the Sun Enterprise Java System. It all comes with the box, and we can decide first whether we want to use it, and if so, whether we want to pay for support.
The second part of this is, I think, a gimme: I can't see running this stuff in production without tying a lot of the risk back to Sun - i.e. if we use it, we're paying.
We could fudge that a bit by using, and therefore paying for, some but not all of it. Personally, however, I just don't feel comfortable breaking Sun's liability for its operation by breaking the integration built into it. In other words, we could save a few bucks on support while probably gaining efficiency and run-time flexibility by replacing bits with more widely used technologies, but if I'm going to use it at all, I'd rather not give Sun an out on any problems that arise.
If we agree that the crown jewels in the Sun software are the identity management stuff and those components aimed at automating multiple systems management then a compromise may seem reasonable: lets dump all of it in favor of more widely supported tools when we're putting the pod into relatively simple environments, and slavishly implement the Sun tools when putting it into more complex environments - even if doing that incurs significant additional costs in terms of support dollars, administrative complexity, and/or loss of efficiency or effectiveness.
Imagine, for example, that the pod replaces some HP-UX servers running Sybase and non of the Java Enterprise components would be needed; imagine it, in contrast, going into an mainframe dominated environment as a CICS/IMS replacement and all of it might have a role.
As a recommendation that idea - stripping the Java enterprise system away for simple environments and holding Sun to everything it promises for complex environments- is based on a fundamental bias on my part: I tend to see the java stuff as perhaps two thirds baked, and increasing in value with the number of servers in the data center - meaning that it's generally not appropriate in situations where your goal is to reduce the number of production servers to one.
So: opinions please?