First thanks for very nice answers. I really learn from these and I like play with this kind of thoughts.
I think TB could loosen up from the ideal fat-tree topology. This would make it a lot more useful and very few users would ever notice any hickups.
I wasn't suggesting that mainstream computers would use dual-TB from their small pci resources.
Better idea would be separate hub, that can be used only when needed. This hub would be topologically at the root level of tree ie. the first thing after the computer. The pipe from hub to computer would be 200% overprovisioned, but I guess that in very rare occassions that would become any hindrance. This hub could have more than 2 ports and nececcary amount of controllers inside it. I'd guess that almost in all user cases, the problems with daisy-chaining is about physical connections, not about bandwidth or lag. In very few cases all TB devices are working with full speed all the time. So there is usually spare bandwidth avaialble.
Other option, if we want to keep strictly to clean fat-tree, would be making a hardware switch. You could attach many devices to a switch and with nice GUI choose which one is "connected". No fiddling with cables.
Hmm,
usb3's awfully long time to mature is not Intel's and Apple fault because:
1) TB is not usb3
2) discreet chips from other vendors were good thing even when they didn't work
3) Apple couldn't helped because they are so small
Sorry, but these just don't cut to me, because
1) 3rd party discreet chips would be needed anyway, even if intel had rushed real working specs through. The difference would be, thet then those chips would work great, now they don't.
2) De facto quirks are needed when specs are not matured. Otherwise you end up with what's happened: you have half dozen different de facto quirks from small companies that don't work together.
Better option would have been de facto quirk from giant like intel and all the small players would have to follow. Just like with usb2 in decade earlier.
Apple could also have used their muscle in this. They still are the biggest and about the only player who makes both hardware and software for their machines. If working usb3 would have been rolled out in 2009 and Apple adopted it in 2010, there would have also not been these false hopes about TB. Apple made the comparison by themselves when offering TB as high speed interconnection and not offering usb3. They were the last manufacturer to offer usb3. The way Apple handled both TB and usb3 shows just how little they care about macs as "state of the art" anymore. Their product philosophy has long been fewer models with better profits and it's only natural that then they consentrate to mainstream products and something like TB means very little to this.
Sad thing here is, that when Apple looses interest in something (xServe, xRAID, macPro, Shake, FCP, Color, even Aperture, maybe even OsX), they don't sell it away, to those who care, who could continue development. They just axe it.
Still the question remains: why usb3 devices are not working with macs as good as with other computers? What Apple has done wrong and why they don't fix it? Is this a way to tell customers that "we told you TB is better, but you didn't believe us..."? Or just that they don't care?
