/sys/doc/ Documentation archive

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: Lucent Technologies & Sun Microsystems

> >Then I realized that the transitor radio seemed silly until you could carry
it
> in a shirt pocket. There are many pluses to networked computing. Like
> reliability (because we can introduce redundancy) 
> 
> What's more inherently reliable than distributed host-free computing on a 
> peer-to-peer LAN??

Well, even though I've been arguing the same point, I'd like to point out that
if you're looking for reliability, none of us should be using software written
after around 1988.  More than a decade after the first Mac, 200MHz PowerMacs
are significantly less stable than what Apple was shipping then.  Likewise for
PCs.

> And how do you figure to institute redundancy on a 
> server-based LAN (especially with something at the client side as brain-dead
as 
> NCs are)?

Redundancy isn't even the major issue.  The real issue is cost. 

You can either buy a relatively inexpensive CPU for each user, and let them do
their computing at their desk, or you can serve them from a machine with a CPU
that's fast enough to serve them all.  Even if you go with max performance =
1/2 users*local CPU, you're paying big money for that compute server, and you
pay more and more $/instruction-second the faster your CPU needs to be above
the mainstream.  Now you make *that* system redundant.

Timesharing sucks when commodity CPUs will do the job just fine, and unless
you're running performance-killing-translation-sloware, they do just fine for
consumer applications.  People who toy around with X, running the occasional
app on a compute server in the basement or on a less loaded machine, didn't
get this lesson hammered home and they're a little too gung-ho about
timesharing (laughable when modern Unix systems wont even support a fraction
of the load, in terms of simultaneous users, with any semblance of the
interactivity that "old" systems did).

I'm guessing we'll see a lot of people suddenly wake up to reality when we see
Oracle completely fail to deliver NCs with performance that's even 1/2 as good
as the inexpensive (sub-$1600) PCs shipping at the time.  No "java vs. java"
doesn't count -- consumers run apps, not languages.  

[...]

> NCs are not likely to be any more "mobile" than a notebook or palmtop PC is.

NCs as most of the NC companies envision them are not only less mobile, they
finally put the hooks into software (machine-keyed licensing, software usage
fees) that PCs in the office were a reaction to.

This is why I love the NC concept.  Anyone who is fed up with the "I've gotta
be a rebel" mindset that is so firmly entrenched in computer guys should
*LOVE* NCs.  

Lots of people are clamoring for it, saying that computers are too expensive,
cheering that we can unseat Wintel (as if that's a worthwhile goal -- unseat
them for *what*?), &c&c&c, and tech people across the industry who can't be
bothered to learn the history of their own industry are jumping up and down
talking about how NCs are going to be Boy Larry's golden spear against the
evil gates.  And they're going to buy, buy, buy into the NC concept.

Here's why it's good: FINALLY we have a chance to get these bozos fired.

And if companies are too stupid to dump these fools after they've tied the
noose themselves by going full-bore into NCs, they deserve them.

RSR