Were barely past the halfway point of 2006, but the most-cited article that eWEEK Labs will produce all year may already be possible to predict. The Labs deep-digging comparison of multiple top-to-bottom IT stacks will please the pragmatists and prickle the purists of both the open-source and proprietary-software persuasions.
Im tempted to eschew summarization in the hope that youll go read the whole thing because much of the value of this work is in the details. Even so, I think I can safely share one top-level conclusion: Dont think that your choice is merely "Windows versus Linux," or even a somewhat less simple ".Net versus Java versus LAMP." Open-source midtier components on a Windows foundation did rather well, and the relative strengths of various stack combinations will present different pictures under different workloads.
This report should inspire you to think differently about the choices available to you, not give you an excuse to stop thinking and just adopt one monolith or another.
Other points also occurred to me as I was looking over these results. First, note the consideration that went into the testing philosophy and the choice of associated tools. The conclusions from the Labs come not from tests to destruction, but from observation of behavior in the course of typical tasks and burdens. Its easier just to load things up until they break, but which would you rather have: a stack thats uniformly a nuisance to configure and maintain, at any workload from negligible to near-infinite, or one thats convenient and well-behaved over a range of workloads that handily covers all of your plausible needs?
Let the prospective buyer beware: If a product is promoted solely on the basis of its capacity, perhaps it falls short in day-to-day operability, and conversely.
I also recalled some of the issues that I raised in my report last year on the broadening scope of the testing that a development shop must be willing and able to address. Its no longer enough to verify that an application, or even a top-to-bottom stack, can do what its supposed to do under the conditions that its builders had in mind. Envisioning the unlikely and defending against the malicious are also part of todays definition of due diligence.
I can imagine that choosing a single-source solution might reduce the number of points of weakness, but I actually lean more in the opposite direction: It seems likely to me that stack components from multiple sources will have interfaces that are (i) more visible and (ii) more likely to be known and understood by the time you get the whole thing working. Well-integrated single-source stacks may offer a simpler out-of-the-box experience to the buyer, but that simplicity may reflect unnecessarily powerful and inadequately defended mechanisms that are silently enabled by default.
Tell me about the stacks youre building at email@example.com.