Re: Linux and the Enterprise (WinNT)

["Robert G. Brown" <bob@xxxxxxx>]

>Thanks for the reference, Bob. This was a very objective-sounding article,
>polite and thoughtful. I'm sure that despite the usual jerking of knees and
>cries of fudd, Linus, Alan and many others will choose to "treat it like a
>challenge!"
We shall see. This is phrased as if it's a serious issue; I'd like to see 
what the kernel korps has to say. As I wrote earlier, this is the same general
kind of discussion as the microkernel vs. monolithic kernel usiness. Linus
seems to be very much against microkernels, and he makes a good case for it
in his section of "Open Sources", and there is an email thread between LT and 
a CS professor/microkernel advocate in the appendix.  Books usually don't
get published with email threads, much less one that's seven years old... so 
imagine that the debate was interesting, to say the least.

>linuxtoday.com has a follow-up from the author, at
>http://linuxtoday.com/stories/5499.html

>where he takes pains to point out that he likes Linux, is trying to help it
>by pointing out its weaknesses so that they can be remedied, and concludes
>with a warning not to trust *any* so-called "independent testers" -- at least
>ones that don't use industry-accepted benchmarks and aren't audited by the
>consortiums who spec those benchmarks.
I read this one carefully, and like many sequels, it's not quite as good as
the first one (one goat to another: "The book was better than the movie"
sorry); The pro-NT bias shows more in this article.

It is interesting how Russinovich pits Linux "against" other UNIX systems and
NT. Since the commericial UNIX vendors have not unified UNIX at all, the pro-
cess of elimination leaves NT and Linux as the only other alternatives, and
Linux is "knocked out" of a "serious competitive position" by a number of 
engineering suppositions, which are not backed up by benchmarks. This borders 
on FUD. Sorry, I liked the original better.

A few things to keep in mind here:

1. The real world is such that only a TINY percentage of the market actually 
   effected by theses arguments in EITHER article, even if the engineering 
   claims are borne out by performance.

2. By the time the technology base involves a lot more than that, the 
   software layer will be different than the one we're looking at now.
   Windows 2000 will be out, and progress will be made on evolving Linux
   SMP features and performance, if needed. The two year horizon is a 
   very long time in this business.

3. Sheer server count (and therefore market/mind- share) go with numbers,
   and there are clearly more departmental/group servers than "Enterprise
   level" systems, at least according to the definitions. Is Russinovich
   therefore abandoning that whole class of systems to Linux?

4. Other issues are not brought up here, such as stability and reliability
   on anything BUT "Enterprise-Level" servers.

Russinovich closed by blasting independent testing labs, and a lot of what
he says here is silly. The ONLY thing labs have to trade on is their integrity.
I do agree that labs often cannot (either due to financial or skill limita-
tions) "max out" all of the products they test, but this also rarely hap-
pens in the real world. One leaves this section of the article, and the 
sequel in general, with the feeling that one has been whined at. Nonetheless,
I reccomend you read this article too, and decide for yourself.

                                                           ---> RGB <---