Randall said:
*cough* It's Moore's Law. It's a tongue-in-cheek
"joke" in hacker/computer jargon.
[Tomb] Randall, just FYI: I teach TCP/IP
programming for Internet applications at our college and I have a background
in both Electronics, Electrical Engineering, and Software Development. I have
a pretty darn good idea of Moore's Law, its origins and whatnot (which is the
underlying basis of my prior comment). I don't go back quite so far as Alan
making PDP's in a kitchen, but I've wirewrapped my own microprocessor system
and written code to drive it in assembler burn onto ROMs. I have a pretty good
idea of who Moore is, what he said, the paper he released on the subject, etc.
And the Jargon File is a questionable resource or reference, despite its
pretensions of grandeur or even adequacy.
Similar to Gate's Law, Parkinson's Law of Data, and Murphy's Law.
[Tomb] My point was that not ONE of these justly
deserves to be called a Law in any rigorous sense.
At best, a half-assed Theorem.
> From dictionary.com, the pertinent parts of the
1) A statement describing a relationship observed to be invariable between or
among phenomena for all cases in which the specified conditions are met: the
law of gravity. 2). In philosophy and physics: A rule of being, operation, or
change, so certain and constant that it is conceived of as imposed by the will
of God or by some controlling authority; as, the law of gravitation; the laws
of motion; the law heredity; the laws of thought; the laws of cause and
effect;
law of self-preservation.
Any of the "alleged" Laws you mention above are at best theorems given the
fact that they cannot be mathematically proven for all cases and that we have
not undergone sufficient time and exhaustive study as to consider them proven.
On the topic particularly of Moore's "Law", there have been many allegations
about how Moore's "Law" will cease to apply in the none too distant future
(there have, fairly, been arguments on the other side). One basis for
attacking Moore's Law is the basic physical limits which will be reached
sometime along the line which limit the minimum size of a switching element
(the core of a processor) based on certain atomic size limitations.
I'm not even saying Moore's "Law" is a terrible
theory - it has held out for longer than people
would have suspected. But to call it a Law is more than vaguely insulting to
real science as it
is number based pseudo-science and
(temporarily supported) conjecture. It is this exact slackness that is part of
what is keeping computer engineering from being properly integrated with the
rest of the engineering profession. (There are other reasons).
> [Tomb] It always takes tech complexity to make
Not true... Robust and simple are OFTEN built in the real world.
[Tomb] I'm more than happy to randomly point to a
hundred instances of an overly complex mechanical or electronic system in my
daily environment.
Feature-rich is a buzzword. It sells. Thus the world
is full of non-simple devices and systems. If it were
otherwise, many modern systems would be far more reliable. I didn't imply that
we never built robust, simple stuff. We just spend a lot of time
making things fancy or feature-rich or snazzy
looking. Often to the detriment of the end product which could have been
better enhanced by more testing, fewer features, more reliability.
The problem is, these terms are _really_
subjective terms.
[Tomb] Conceded. However, in the sense of a
colonial piece of equipment, there is a clear absolute standard: how does it
do its job, and how long does it last.
Simple and robust chips are much more complex
and unreliable when compared to fire-starting
equipment like a lighter.
[Tomb] Even this statement is problematic. I think
modern computer memory could be considered simple, but these will last 15
years? Will the lighter? And aren't you comparing apples and oranges? Wouldn't
it make more sense to compare
a zippo to a modern piezo-electric lighter with an
element that can put out 2000 degrees? Or to compare a 386 chip rated for the
space program (simple, in this comparison, and robust) to a bleeding edge AMD
chip? Which is more robust?
Which is much more complex than say a match.
Most first run products are simple, feature-poor
commodities. It's the 2.0 and beyond that get nasty.
[Tomb] In an economic sense, most prototypes are
either underengineered or overengineered (probably in equal measure or a bit
of both). As time goes by, things get made lighter (sometimes
at the cost of durability - who needs a 286 to last
20 years? We throw it out in 5), sometimes features get added. Sometimes later
models are refined with the edges taken off (leatherman tools
are an example) and sometimes they are feature-
heavy bloaters with dubious reliability.
Simple and robust are not always easier to manufacture either. It is a simple
procedure you can do in the house to create Oxygen and Hydrogen, (both fairly
unstable in this
condition). Many other forms of fire-starting
equipment require massive refineries, complex chemical processes, etc. Still,
when it comes down to it... I'd rather a match to start my fire.
[Tomb] The match is a fairly simple construct.
Simple, robust, and does the job well. Of course, so might a zippo. Robustness
is not always easy to engineer into an object. Part of this challenge is
dependent on the nature of the object and its function. Simplicity can be hard
to engineer in if the task it must perform is complex. But the absence of
simplicity introduces plenty of room for mistakes and plenty of places for
failures. Complexity is the enemy of Robustness. This is ultimately the basic
lesson of Reliability Engineering. Herein we learn that, for systems of high
reliability (0.97 or better I recall), the optimal number of redudant parallel
switched backups is 2-3. Beyond that, the
complexity you are introducing actually reduces your overall reliability.
Complexity and fancy neato gee whiz features sell new cars in a showroom on
Albion. The ability for a binary propellant rifle to function reliably when
submersed in muck is much more of a selling feature on Slimobia III.
Tomb
Since you seem to want credentials, I'll spout a few of mine... (DSW are so
much fun.)
Senior Systems Engineer for CNN. My (and my teams) duties include the
hardware, OS, software, networking, and every little piece of everything
inbetween. I'm responsible (in part) for over 600 systems, 300 peripherals,
etc. etc. etc. I have more certifications than I care to remember. I double
majored in Comp Sci and Chem. I've done professional developing of complex
systems, and am proficient
(maybe not speedy or fresh and not current in most) in over 10+ separate
languages, 25+ variations of those 10+ languages.
I've been in the real world of IT for many years. I've been on the net for
longer. I still remember the 3 heirarchies of usenet pre-great
renaming.
Damn, I feel like I'm applying for a job.
I don't see where your phallus is any bigger than mine. And if you can take
some ribbing meant in good nature, I'll remind you of the cliche, "Those you
can't, teach." (If it you can't, know that I'm probably sleeping on my couch
tonight for having said that... The better half is a teacher.)
While I concede that the Jargon file is not a great source, it is kept as
factual as possible, and is one of _the_ sources of hacker cultural
history.
Now, with preliminaries and pleasantries out of the way, I'll repeat myself.
"Moore's Law. It's a tongue-in-cheek 'joke' in hacker/computer jargon"
When referenced as Moore's Law, it's meant to be taken exactly the same as
Murphy's Law. They are meant to poke fun at many things science being but one
of them, much like most hacker humor, and also like most hacker humor, they
have a gain of truth in them, and again, like most hacker humor they
are somewhat good-natured cynicism. Not one of those I mentioned is
meant to be taken in a rigorous fashion. At best, when being serious, they're
meant to be commentary on current (and past) societal problems and events. It
seems you didn't really look those other laws up. Gate's is
really amusing in a cynical sort of way, given the software worlds trend to
bloating.
As to your definition of law, you've forgotten 2 very common and important
types of law. Socio-political, and religious. Neither of which needs
to be proved, and in fact most can't. (I admit to being geek enough to have
spent time amusing my self this evening by trying to write proofs of "Thou
shalt not commit covet thy neighbors wife" and several US gun laws. Failing,
but amusingly.) I won't pretend that Moore's law is really religious, but
there is a hint of it.
Oddly enough, dictionary.com responds to a search of Moore's Law by giving
quotes from the Jargon file but has no entries for Moore's theorum.
On the topic of Moore's theorum, I'll give you your statements... With the
exception that the theorum as first presented is certainly going to be proved
dead in a very short time. None of Quatum computing,
Gated/switched
light (fiber-optic) computing, nano, nor very very recent
chip-manufactoring techniques involve silicon, which is directly stated
in the theorum. "...the logic density of silicon integrated circuits..."
The general theorum is applied too liberally, or misapplied due to a
misunderstanding (or misuse) of the above Moore's Law. However, the general
theorum, computers "double" every 18 months IS provable as a law in the second
definition of your limited definitions. It's really a limited and specific
case of evolution when applied to technology.
[Tomb] I'm more than happy to randomly point to a
> hundred instances of an overly complex mechanical
Come on... I can point to as many examples of simple mechanical and electrical
systems in my daily environment as you can point to complex. I'll start by
pointing out the fact that complex systems are
_OFTEN_ built of simple systems. Which should be enough to prove my
comment that "Robust and simple are OFTEN built in the real world."
You're right... We do spend a lot of time adding features. By that statement,
2.0 is more complex than 1.0, correct? But that means 1.0 is
simple in comparison... My point made. Simplicity is relative.
Modern memory, if treated right, will last much more than 15 years... Frankly,
I'd be surprised if it didn't last darn near
forever. Memory does not have _any_ planned obsolescence. Or, more to
the point, any and all planned obsolescence is external to the memory. It's
actually a fairly unique time in the history of manufacturing. Computers in
general, have no need for planned obsolescence. The market, technological
growth, new tech vectors, and other factors give all the necessary
obsolescence needed for the majority of computer
parts. Hard-drives and monitors are 2 exceptions, and you will note
that they have planned obsolescence built into them. And it makes sense...
Hard drives can be used from computer to computer, and thus the impetus to
purchase another is low, yet RAM can be changed simply by changing the socket
it's on, or increasing the need of it.
Some trench practical knowledge: Any ram you get, if properly burned in
before use, will last much longer than the life of the machine. Failure
rates are 5 to 6 nines or better. Burn in is required only to isolate chips
that are defective. How's that for robust? Look at the difference between a PC
and a 5 9 reliability machine like Sun's high end E class machines. They use
similar or the same ram. Believe me, if Sun claims 5 9's reliability, they
mean it. The need (and price) of such reliability
means that anything requiring it is extremely valuable or dangerous, or both.
Litterally, the cost of a 1 minute downtime in that situation can
mean millions of lives, or billions of dollars lost.
I'd be surprised if any firearm, not treated properly, would last even 15
years. For that matter, most anything, if abused, tends to not last long. With
or without planned obsolescence.
Believe it or not, aside from the odd bug, the most bleeding edge CPU's are as
reliable (if operated correctly) as 8086's, 8088's, 286's, 386's, etc. They're
even as robust. Run a modern chip at 8088 speeds, and you'll still see an
increase in performance, for the same or less heat and physical resistance.
When I said first run, I was not speaking of prototypes. As to your
over/under engineered comment, I disagree. Most first run objects are
engineered correctly (I will _NOT_ include modern software here however)
to the specs. It's almost always the fault of bad specs, which is usually the
fault of bad market research, miscommunication, artifical limitations placed
upon engineering, misunderstood needs, unforeseen problem resulting in product
use, etc. True, this doesn't lead to robust and simple machines, but that is
not the engineer's fault.
I do agree with your complex does not make robust or reliable easy. But it
does not preclude it either.
*sigh* Rand.
> At 02:58 AM 1/31/02 -0500, Thomas Barclay wrote: