Care to elaborate? Cisco, Lucent and other vendors have used this approach for years to great effect. Ditto, music hardware vendors.
Apple isn't going to retreat all the way back to the XServe to re-imagine the Mac Pro. The more likely reasonable expectation is that Apple won't make something that isn't overly rack hostile. That is about as close to "rack oriented' they are probably going to get.
The aren't looking to fit into some data center rack or some rack-on-wheels. If it happens to fit; great. but it isn't going to be a primary objective. A data center room isn't a good place to go looking for standard Apple design cues. Apple probably will do a some more limited size reduction (relative to the 2006-2012 baseline dimensions) that will "happen to work" to make it less rack hostile, but not really rack friendly.
The standard I/O sockets will be fixed. Just like the previous Mac Pros ( and other Macs.)
For instance, you buy the base chassis with one CPU module and X amount of DRAM slots, another module with a GPU (AMD or NVIDIA), a module with PCIe slots for your specialized cards (Red Rocket, SAS, RAID, PCIe storage), and have room to add a second CPU module with DRAM slots,
Some variation on some blade server chassis. Not likely at all. Didn't even follow that in XServe heydays.
...it satisfies the LEGO bricks rumors and allows the largest amount of flexibility to the end user. Pros don’t care about the aesthetics, as has been stated here, so a clean front panel covers it all up.
Apple has never even remotely hinted directly that they are on the LEGO brick track. Their explicit prototypical example of modularity about external Displays. That is more than significantly far from the LEGO brick model.
I won’t pretend to understand the engineering issues that might present themselves, so maybe that is what took Apple so long.
I think folks who keep piling up the engineering complexity into an every taller stack to excuse the long delay are going to be deeply disappointed. This is probably not gonig to be some "Moonshot" / "invent warp drive" kind of system. Apple will have stuck some requirements some folks may not have on the system but most lkley it is going to be fairly straight forward in what they need to do. Apple probably didn't start full time resources allocation until 2017 time frame so much of the gap was spent doing not much that will have direct, specific impact on this system development time.
Couple that with a few "bet the farm" on delinquent components and that is probably the main bulk of the delay time.
So, tell me what is so horrible about this approach? Considering it is really just the old cheesegrater approach taken just a bit further.
1. A substantive number of folks who will buy a Mac Pro aren't going to put it into a rack. Under their desk. Some one their desk, but physically managed is rigid container is probably not the majority of the installs. Apple makes those deployment hard that will dampen sales.
2. Apple has very little interest in a BTO options page that scrolls on for "days' with every gewgaw imaginable. Overly complicated inventory probably means more folks inside of Apple itching to kill it. For a "nice to have" product as opposed to a "need to have" product that is not a good position to be in.
3 Since having a very limited amount of copious spare time has probably greatly contributed to the long upgrade cycle for the Mac Pro, a series of custom models just means more work. More work would lead to more time . Yet another super long refresh and probably going to get " death spiral product".
Apple needs something easier to keep up. That doesn't means they are simply going to run down to Fry's for parts update details. Something functional but not too complicated for the basics. In some limited incremenal augments more standard.
4. The decoupling of major subsystems like GPU and CPU from RAM really doesn't work well. It really doesn't buy much in cost effectiveness either.
