Log in

No account? Create an account
Pete Zaitcev's Journal -- Day [entries|friends|calendar]
Pete Zaitcev

[ userinfo | livejournal userinfo ]
[ calendar | livejournal calendar ]

[09 Oct 2004|11:09am]
Mindless Elitism Attempts to Claim Another Victim, Victim Resists.

Tim Blair writes: "Never again will I doubt the wisdom of Australians." Let us see if he will doubt the wisdom of Americans now!

Wit aside, so much for the punditry. I had my doubts as well, only I did not try to broadcast them to the world as a gospel.

But the big lesson is for Canadians. It remains to be seen if they accept it. I have my doubts. Take this, Tim. For a healthy rational mind, doubts are tools. Do not discard them so quickly.
post comment

[09 Oct 2004|02:46pm]
Regarding the recent Sun hoopla with belligerent Scott McNealy, I cannot help but to look back to my Sun tenure. It was a big experience, which left its mark. Sometimes it even calls me to work on Microsoft for a few years, just to see how it is. At one point I noticed that often I speak to other Redhatters and say, "when I worked at Sun, we did this and that, therefore...". For example, Sun has, or used to have, different tracking systems for customer problems and engineering bugs, which makes a lot of sense. I do not know if late Steve Taylor or others thought about it when Issue-tracker was conceived, or it just happened. But calls are heard often about merging Issue-tracker and Bugzilla. I always confront those. Anyway, when I became aware of the habit I decided not to mention Sun's practices again. Majority of them were common sense anyway, and we discovered those I knew independently. We have our own ON gatekeeper(s) now, for instance, and it happened without my input whatsoever.

We also discovered same limitations, too. For example, the rate of change in the centralized model must be limited to prevent the whole product from falling apart. Taking just the kernel (which is bigger than the good ole ON now), the team is so big these days that I do not even know many (primarily new hires and former Sistina folks). And yet all those people cannot do one hundredths of the development Linux at large sustains. Strictly speaking, Linux is limited. Only Ernie's rate limiter is set to 1/100th of Andrew's. So, it's almost as if Linux moved at an infinite speed.

Looking back at Sun from this vantage, Solaris could compete successfully with Red Hat Enterprise Linux, if Red Hat were the only one to do the development, but we have an ultimate weapon here. It is called "upstream". Once the upstream is accounted, I do not see how Sun can even dream about any parity. It's a pure delusion.

The one thing they could do is to have a few really good people (like Bonwick or Henson (never seen Cantrill's code, so I cannot tell)), and try to identify a few very narrow improvements with wide ranged implications, such as SLAB or Dprobes, implement those, and try to ride it out. It would also be essential to patent those to deny those improvements to competitors. Did I mention that ZFS must be patented to hell and back for my guess being valid?

Having a few key advantages might not be enough. Firstly, we do not know if enough of their customers are single-issue buyers. Linux is broader, and it's not shallow either, it's just deeper in other places. But forget this. I actually doubt that Sun can even execute the modest proposal successfully, because one of the defining characteristics of their corporate culture was debilitating bureaucracy. And PSARC was not the worst of it. I kind of liked PSARC. Rather, the whole atmosphere was stifling. Just about nothing could be done without approvals at the top.

For a long time company engineers were simply in denial about the corruptive effect the bureaucratic culture had on development and tried to joke it off. They were quite open about the imperative to use unrelated project funds to sustain next generation projects, for example. But they cannot isolate themselves completely. Some even accept the machine. I blogged once how Karen Clark came down on me for daring to download her precious SunCluster source even though it was related to the work I did. What if I were simply curious? Would she try to get me fired?

Having seen all that I'm not going to keep quiet when Red Hat does it. And we have a few people with what I perceive a similar spirit. Matthew's leadership is nothing like McNealy either. So, I'm optimistic. If Scott comes out swinging, then he's asking for what happens.
post comment

[09 Oct 2004|07:57pm]
I pushed my luck quite a bit today by purchasing a VIA C3 based mini-ITX box as my new home server. I wish to keep the local power dissipation in check, because my room is very small. I use a 2.5" drive in it. In the event, the C3 turned out to be a Samuel 2. Using newer kernels which have Ingo's fix for the PMD base calculation I installed Fedora Core 2 on the little thing. So far so good. Running "yum update" now.

I wish Transmeta or AMD made a CPU for little servers with low power dissipation which sold for $203.00 (before tax), because VIA is, frankly, an insane vendor. They have a long history of semi-broken CPUs (such as mysterious lockups of Erza cores), never publishing any errata, and of course never explaining what the heck is actually happening with gazillions of cores they supply to the market. But so far, there is just no other way but to buy VIA.

Other than the CPU shenanigans, the box looks good. I opted for a normal power supply, as I am getting irritated with invasion of little black bricks, even including those which do not use 3 power outlets at once (by plugging into one and obscuring two more). This meant a fan, but hey... At the price I was paying it's hard to complain.
post comment

[ viewing | October 9th, 2004 ]
[ go | previous day|next day ]