Please note that I have not had a chance to see and use the new MacOS X "Leopard" server release. Comments specific to Leopard Server here are, therefore, based on third party reports.
It appears that the big changes in the server version focus on ease of use, particularly with respect to network setup, and on getting along in a predominantly Windows environment.
As a result the new MacOS X Server is much more competitive with Linux than "Tiger" was - and especially so for very small businesses with little or no professional IT support and the courage to look beyond Microsoft.
When you compare MacOS X to Microsoft's servers the big issues are reliability, ease of installation, licensing cost, and operational complexity - with licensing cost the most important for small businesses because these tend to focus on initial capital costs. Thus Apple's decision to include free ecommunications software with the server coupled with its inability to charge client licensing makes MacOS X increasingly attractive as the number of clients grows.
When you compare MacOS X server to Linux, however, the key advantage for small businesses isn't capital cost, it's ease of setup and use. Fundamentally that comes down to the issue of how the small business gets its servers to work: because the difference between clicking through a GUIfied process and calling a script is enormously valuable if you don't know what scripts to call, and pretty much valueless if you do.
In other words, Apple's ease of use advantage over Linux depends ultimately on an impersonal sales and deployment model: if joe average small business guy gets his Intel server delivered by FedEx and then has to set it up himself, Apple's software can really help - but if Jane's Computers hands it over it all nicely set-up and running, then it will take long term customer behavioral change before the customer sees any benefit from Apple's systems management software.
For larger businesses, neither licensing costs nor set-up costs are likely to drive the decision. Instead, at least in most cases, what's likely to drive the decision is a a basic difference in philosophy between the Mac and Linux camps.
That difference comes down to this: the Linux focus is on replacing other technologies, and particularly those from Microsoft, where Apple's focus is on using its server software to make it easier for MacOS X desktop users to co-exist in Wintel dominated environments.
Thus Linux scores its greatest successes against Microsoft by competing for data center dollars and accommodates Wintel clients as part of the cost of doing that. Apple, in contrast, competes best with Microsoft on the desktop and accommodates Wintel servers as part of the cost of doing that. Think of the Linux desktop as a parallel to Apple's data center servers and you get the superficial part of the picture. Look deeper, however, and what you see is a key difference: Apple uses its servers as part of a viral marketing campaign aimed at making it easier for minority Mac users to live, work, and proselytize in Wintel corporate environments - but the Linux community doesn't use its server room advantages to push the Linux desktop.
This would seem to be as much an artifact of community structure as anything else but the deeper roots of this come down to a difference in philosophy: the Mac started as a better desktop, Linux as a free x86 Unix - and both continue those traditions. Thus Apple's latest server technologies are all aimed at helping desktop users work in Wintel environments but no such co-existence strategy exists on the Linux side - indeed the general Linux approach to the data center is closer to that of science fiction's homogenizing swarm than that of the symbiote.
To a wintel manager facing user demand for Macs, Apple's strategy produces both fires and frying pans: fires because "Leopard" makes it even harder to argue that letting Macs into the environment will disrupt the smooth functioning (!) of the Wintel networks and servers in place - and frying pans because the new features make it much easier for users to silently augment Windows data center servers for file, print, and authentication functions with their own machines running MacOS X.
Nowhere is this difference clearer than in the murky world of "identity management" where Linux is clearly about Windows services replacement while Apple's multi-technology mash-up is equally clearly about protecting Windows services in Mac desktop deployments.
Unfortunately Apple's co-existence strategy leads to a major problem in that it makes Apple a follower - forced to play catch up whenever Microsoft changes its technologies. In other words, companies which try to take advantage of Apple's strategy will eventually find themselves trying to hold off both their internal wintel enthusiasts and Microsoft's stealth upgrades while waiting for Apple - and because that won't work they'll suffer the usual consequences in terms of organizational disruption and mutual finger pointing.
That's why my bottom line on MacOS X Server versus Linux is simple: ordinary users don't have a choice, if wintel is an organizational given then Apple's co-existence strategy makes MacOS X preferable to Linux, but for anything bigger than a mom and pop shop, the right data center solution is Linux, not MacOS X. Why? tecause when it comes to cleaning the Augean stables front end loaders just work better than perfume spritzers.