If you didn't catch the Windows 7 kernel debate between OSNews' Thom Holwerda and InfoWorld's Randall Kennedy, I highly recommend it. Basically, via a series of back-and-forth emails, the two agreed to debate Kennedy's claim that because the Windows 7 kernel has a similar thread count to Windows Vista, it is a minor point-type release and probably will have Vista-like performance.
Personally, I think Holwerda had the more logical argument. Simply looking at the number of running threads won't tell you whether those threads differ from threads in previous versions and if performance is directly proportional to thread count. Granted, history might reveal a common trend with thread count and performance, but that's still a big assumption.
Kennedy acknowledges that he's drawing conclusions, but feels it's warranted based on his knowledge of kernel changes over the years. I side with Holwerda on the issue, but I don't have a problem with Kennedy looking at the thread count metric and drawing some conclusions--even if they are a little hasty. What I do have a problem with is his motive behind looking at metrics like the thread count.
The first thing I noticed in the debate was Kennedy's skepticism around Windows 7. I don't know if the "Mojave campaign" raised his suspicions or what, but Kennedy immediately came across as doubtful about Windows 7's alleged improvements from the start. I couldn't help thinking that he was chomping at the bit to get his hands on beta code to prove himself right. Furthermore, I'm trying to figure out what Kennedy was expecting. Take the following example from his benchmarking, which prompted the whole debate:
In fact, as I worked my way through the process lists of the two operating systems, I was struck by the extent of the similarities.
Really? Was he thinking that Microsoft spent tons of time and money re-architecting the code base for Longhorn so engineers could go back to the drawing and start all over from scratch using the XP code base? Yes, Vista had some problems early on and wasn't well-received, but that's due to a combination of factors--not necessarily problems with the kernel. In fact, consider the following comment from eWeek regarding Windows Server 2008:
And, yet, Microsoft's Windows Server 2008, which recently hit the RTM (release to manufacturing) milestone, demonstrates that Microsoft is capable of producing a lean, mean server machine—and doing it, no less, atop the same code base that backs the company's oft-maligned Windows Vista client operating system.
Well, what do you know: the same code base used in Vista was used in the "lean, mean" Windows Server 2008. To me, that would suggest that Vista's alleged "problems" likely are rooted in the user-land or other aspects of the OS--not the code base. In fact, Windows Server 2008 seems largely well-received. Sure, uptake may be relatively lukewarm and methodical as most server upgrades are, but customers are pleased with the security, usability, performance and functionality enhancements. Several IT shops I've talked to already are testing the server OS and planning an upgrade plan for 2009. So, the code base is good enough for them, but apparently not good enough for Kennedy.
Oddly, though Kennedy recalls Microsoft's public suggestions that Windows 7 would be based on the same Longhorn platform as Vista/Server 2008, he doesn't set his expectations accordingly, as evident in the following example (I've included the section heading as well):
Lipstick on the pig
So where does this leave us? For starters, we can now say with some certainty that Windows 7 is in fact just a repackaging of Windows Vista -- an "R2" release, to use Microsoft's nomenclature on the Windows Server side of the house.
So Kennedy doesn't mind making assumptions about kernel thread count and performance, but when Microsoft announced that the server-side name for Windows 7 would be Windows Server 2008 R2, he didn't want to assume that both upcoming OSes would be based on the same existing platform. Nope, he was expecting MS to reinvent the wheel even though the "pig's" Longhorn platform was looking pretty darn good on the server side.
Like I said earlier, I don't mind a speculative review or benchmarking of pre-release code as long as it's honest and objective. I'm not sure where along his 16-year history with the Windows kernel he got so jaded, but Kennedy obviously has an agenda here. I could highlight even more examples like those mentioned above, but I think you get the idea: this is FUD designed to scare you into thinking that Windows 7 will largely be more of the same (wait, that sounds familiar... ;-). In fact, Kennedy triumphantly announces: "Welcome to Windows Vista R2!"
<sarcasm> Uh oh, watch out, Microsoft: the cat's out of the bag! </sarcasm> I'm sorry, Randall, that you already are so disappointed in Windows 7 based on pre-release code. But really, I think you wanted to be disappointed and would've looked for some justification for it in any code you tested.