Linux-Advocacy Digest #875, Volume #27 Sat, 22 Jul 00 16:13:05 EDT
Contents:
Re: Linsux as a desktop platform (T. Max Devlin)
Re: Mrs Drestin Black (Jacques Guy)
Re: Linsux as a desktop platform (T. Max Devlin)
Re: What I've always said: Netcraft numbers of full of it (mlw)
Re: Linsux as a desktop platform (Gary Hallock)
Re: Linsux as a desktop platform (T. Max Devlin)
Re: If Microsoft starts renting apts (was: If Microsoft starts renting apps) (T. Max
Devlin)
Re: Richard Stallman's Politics (was: Linux is awesome! (Graham Murray)
----------------------------------------------------------------------------
From: T. Max Devlin <[EMAIL PROTECTED]>
Crossposted-To: comp.sys.mac.advocacy,comp.os.ms-windows.advocacy,comp.unix.advocacy
Subject: Re: Linsux as a desktop platform
Date: Sat, 22 Jul 2000 15:00:05 -0400
Reply-To: [EMAIL PROTECTED]
Said John W. Stevens in comp.os.linux.advocacy;
>"T. Max Devlin" wrote:
>>
>> You are missing my point.
>
>Max, consider the possibility that your point was badly stated.
Quite possible. If you have any suggestions, I'd be happy to hear them.
I think it may be so contrary to your experience that you might consider
any statement of it to be difficult to understand.
>I, too, consider what you wrote to be unreasonable, or ill-informed.
>
>Please re-state your point.
>
>> But that doesn't change the fact that you are missing my point, which is
>> that optimizing the system for whatever task the user is currently
>> engaged in may very well be more efficient than not optimizing the
>> system for whatever task the user is currently interfacing with, but
>> only for the average of all potential tasks.
>
>This assumes that running a game of solitaire is more efficient than
>rendering your 3D image . . . the task you get paid for, while playing
>solitaire is *NOT* something you get paid for.
No, it assumes that rendering the 3D image would be so greatly optimized
that you wouldn't have time to play solitaire while you wait. ;-)
>The blind assumption that the interactive task should be the highest
>priority task at all times is just plain wrong.
Again, I think this resolves on a preconception of a singular task,
which I didn't take for granted. I would suggest that it could be
reasonable to call any system which this statement applies to a
"workstation". On a desktop, not only is the interactive task the most
important, it is often the only end-user task to speak of.
>The highest priority task should always be that task that maximizes the
>value of your CPU cycle investment.
The priority of tasks is entirely independent, and certainly not
determined, by CPU cycles. Tasks have priority for reasons end-users
(and their bosses, often) determine, not CPUs and engineering
efficiencies.
>> Not the *process*, the
>> *task*.
>
>Stop. Before throwing such words around, you need to be very sure you
>understand what they mean. What, in your mind, do the words "task" and
>"process" mean? For a computer scientist, a process and a task are
>differentiated primarily by which can contain the other.
By process, I refer to the same concept, I believe, as the CS field. I
didn't think that 'task' actually has a rigorous definition, even in CS.
The 'task' is the higher order abstraction of 'why you're running the
process'. While CS may recognize that a task can "contain" a process, I
believe that is simply a single order abstraction because that is all
that the engineering disciplines are concerned with. What I am thinking
of as a task is more of a "gestalt" of the ultimate purpose for using
the computer, not a single step in a work procedure. I'm not sure what
alternatives might be more appropriate, if there is a specific conflict.
>> The PMT issue is clearly a matter of balancing load with
>> responsiveness.
>
>No, it's not. Here is where your understanding of PMT fails. PMT
>increases the value of your CPU cycle investment.
I can't help but think that you are starting with an assumption that I
must be wrong, and further assuming that all those things I say which
you didn't understand simply don't make sense to begin with. These are
not necessarily incorrect assumptions, apart from the fact that all
assumptions are by nature incorrect.
I was provided in no uncertain terms with an explanation of PMT which
was almost literally "a matter of balancing load with responsiveness",
which is to say "how to keep the system responsive under load". If you
want to re-phrase it, feel free. But give me a reason for that beyond
your intent to show I don't "understand" something.
[...]
>And your viewpoint is flawed. Hire an engineer to engineer a system
>specifically for your "individual reasons", and you'll get a system
>optimized for your "individual reasons".
I think you have the wrong image of what I'm describing. When
considering "average versus peak", I'm talking an instantaneous set of
tasks' and processes' requirements, not all possible processes and
tasks.
A PMT system attempts to provide a level of service, if you will, to all
processes. But other than kernel/user and CPU-I/O bound, it hasn't any
more awareness of the actual characteristics or purposes or
implementations of the tasks, the applications, the programs, or the
processes. My idea was one where the system can "know" whether to
starve the renderer or not, because sometimes that makes the system more
efficient at something of a higher "real" priority for the end user.
>To blame "engineers" for the results of *YOUR* choice to use a general
>purpose system to perform your specific tasks is to blame the wrong
>person. Go look in the mirror, and talk to that person about what a
>poor choice they made.
I don't blame anyone for anything. I am speaking entirely of general
purpose computing, and the fact that the same scheduling method is
considered optimum at all points in time and for all sets of processes
and tasks.
>> Certainly a secondary concern in comparison to the
>> design details, but it disconcerts me to think that it is not even
>> considered as remotely important, and even silly.
>
>What is silly, is to believe that these factors are not even
>considered. OF COURSE they are considered! But when you build a system
>whose design purpose is to provide a generic, somewhat abstract service,
>you cannot focus on the needs of a single user. You can build the
>system to be flexible and configurable enough that the individual can
>tune the system somewhat, but to build an entire system just for Max, is
>a very, very expensive proposition. I estimate that a fully optimized
>system for you alone would cost from between 1.5 and 3 billion dollars.
>
>Write me a check, and I'll be happy to start the process. . .
If you considered my thoughts to be ludicrous, you misunderstood them,
plain and simple. They may not be incorrect, they may not even be
worthwhile, but they are not ludicrous; they're ideas. If you can only
deal with them through ridicule, then apparently you aren't up to the
challenge of considering them thoroughly enough or addressing them
directly.
*Negatively* impacting the *average* 'efficiency' of a system in order
to increase the particular performance of the overall system might be a
more efficient mechanism for the "generally used for one task at a time"
desktop PC, and might not, require the expense of a special-purpose
system, as you grandly but accurately suggested. A PC capable of
running any set of tasks might be "self-optimizing", for whatever set of
tasks are currently in use. How this would be implemented or whether it
would be efficient I must leave to those with both the engineering
knowledge to ponder it and the ability to understand it.
You have side-swiped my original intent, I think. I don't need "fully
optimized" for 2 billion dollars. Can I get "80% optimized" for $2000?
I don't want to have to tune it; can't the computer do that
automatically? PMT is certainly a very powerful way to run scheduling.
I'm not the first to suggest, however, that something even better may be
possible.
--
T. Max Devlin
Manager of Research & Educational Services
Managed Services
ELTRAX Technology Services Group
[EMAIL PROTECTED]
-[Opinions expressed are my own; everyone else, including
my employer, has to pay for them, subject to
applicable licensing agreement]-
====== Posted via Newsfeeds.Com, Uncensored Usenet News ======
http://www.newsfeeds.com - The #1 Newsgroup Service in the World!
======= Over 80,000 Newsgroups = 16 Different Servers! ======
------------------------------
Date: Sat, 22 Jul 2000 19:24:51 +0000
From: Jacques Guy <[EMAIL PROTECTED]>
Subject: Re: Mrs Drestin Black
abraxas wrote:
> But he certianly is not a programmer of any sort, and very probably
> not even an 'IT' professional.
Nor a Web designer. If you tune in on his Osin Wavz (no, Tim Palmer
was NOT here), on his Osin Wavz's page, you'll be greeted by a
107K GIF of something that might take 10K in JPEG. And if you tune
in onto his other paramour's site
(http://www.beckysunshine.com/bsunfrm.htm)
you'll see that *ALL* the pix of the buxom wench are GIFs. JPEG
probably
sucks like Linux out there...
------------------------------
From: T. Max Devlin <[EMAIL PROTECTED]>
Crossposted-To: comp.sys.mac.advocacy,comp.os.ms-windows.advocacy,comp.unix.advocacy
Subject: Re: Linsux as a desktop platform
Date: Sat, 22 Jul 2000 15:26:55 -0400
Reply-To: [EMAIL PROTECTED]
Said John W. Stevens in comp.os.linux.advocacy;
>"T. Max Devlin" wrote:
>> Said John W. Stevens in comp.os.linux.advocacy;
[...]
>> >Just because a car is just a car, does not mean that it only has one
>> >part.
>>
>> But I can't be driving two cars at the same time.
>
>Ok, obviously that analogy didn't work for you . . .
Car analogies rarely do. ;-)
>> I can't be "using"
>> two applications at the same time, either.
>
>Irrelevant. "Applications" are not automatically and in all cases one
>process.
Just because it is not a deterministic relationship does not mean it is
irrelevant. That is certainly a large part of my point, I think. To
the scheduler, all processes are "equal", so to speak, though each might
have priorities and nice values to attempt to present their "importance"
to the scheduler. Still, the scheduler has no awareness of which
processes are linked by being part of the same application (now, or in
general). Would there be no extra performance to be gained by paying
attention to this, so that if one process of an app is a bottleneck, the
other processes in that app are handled in whatever way might reduce the
bottleneck?
>> Obviously the value of
>> multi-tasking, of having more than one application *available* for
>> task-switching, and actively processing while in the background, cannot
>> be over-estimated. But it is not the issue I was addressing.
>
>You missed the point: a single (modern) application *IS* multiple tasks.
Yes, you've already noted that the term 'task' is problematic. Have you
a suggestion for what to call something which includes (potentially)
multiple applications? The end user task was my focus, not the
"collection of processes" task of the software designer. Hopelessly
vague, and changing definition instantaneously and quickly, the "higher
order" task is less easily quantified, but far more important in the
end, than *any* individual 'task' the computer is performing.
>> >A word processor might spawn off as many as 31 different, separate
>> >threads of execution.
>>
>> I did not say there is only one thread at a time, nor only one process
>> at a time. Merely that there is only one task being performed by the
>> user at any one time.
>
>And that is irrelevant to the point being discussed. How many
>"applications" the user sees running is a virtualization (IOW: a
>computer generated lie) that may hide as many as 31 separate processes
>(been there, counted 'em up).
>
>Add in the system processes, and you end up with a fair amount of stuff
>going on.
Precisely my point. Why is scheduling method that optimizes the use of
the CPU by multiple processes automatically considered to optimize the
CPU for multiple *tasks*, or *applications*, or *tasks*? Perhaps
something less than optimal for processes would be more efficient for
the user's purposes, even if it seems an inefficient waste from the
engineer's more technical perspective?
We get back, again, to Ethernet. Which *sucks* at loads of greater than
30%. But is *very* efficient at 10% utilization. Even more efficient,
in fact, than any system which doesn't get bogged down at higher loads.
But it is widely recognized that a CPU is *rarely* a critically limited
resource given typical desktop use. Is it not possible that a
multi-tasking system which is "optimized" for desktop use rather than
workstation or server use might have a similar result? Getting the
computer to 'be faster' when the average load is 2%-5% might be more
advantageous in such an environment than one that doesn't slow down
horribly when used more aggressively. And given the local nature and
cognizance of the system, it could switch to the more egalitarian method
when the load does run higher.
>> That task may indeed be a combination of
>> procedures (watching a download's progress while playing a game or
>> writing a note while a database sorts),
>
>NO Max! A *SINGLE* application. Not an FTP client *AND* a game program
>*AND* a data base system, a *SINGLE* word processing application may be,
>under the pretty lies, *31* separate tasks.
Sorry for treading on holy nomenclature. Couldn't you maybe consider
that the term 'task', for all its engineering value, is not a term
strictly and solely defined within the world of engineering? It seems
obvious, as well as regularly remarked, that I'm not an engineer. So
why you are insisting that my statement don't make sense using the
engineer's definition of 'task', when obviously I'm not using any such
strict formal concept, but merely the routine definition of the word
'task': work which a person is faced with performing.
>Consider "real time spell and grammar checking", just for starters . . .
>and then turn on that @$# paper clip . . . no, don't do that. I'm
>sorry, that was cruel . . . :-)
>
>> but from the user's perspective,
>> it is a single conglomerate task with a single desktop interface.
>
>If the issue under discussion was solely: User Interfaces, this would be
>relevant to the discussion. But when discussing how various types of MT
>affect the user experience, the total number of applications the user
>*THINKS* are running is irelevant, so long as the method chosen for
>switching between the various threads of execution allows the
>virtualization to "work".
Which is why I stated essentially the same thing when I pointed out that
even if multiple applications are in use, to the user it is a matter of
one 'task' they are performing. How many 'tasks' or whatever the
computer is performing is quite beside the point, other than to note
that it is optimizing the computer for the *users* task which is
important, not optimizing the computer for the minimal CPU bottlenecks.
In most cases, the two are handled well by PMT. I would prefer if, when
things do get bogged down, the user is given preference, if possible,
over the putative average value of the CPU scheduling mechanism.
[...]
>> >And, just as importantly, there are a number of tasks running on your
>> >average workstation even when *you*, the *user*, hasn't started any
>> >application.
>>
>> Why is it that everyone thinks I'm trying to hand-wave task-switching
>> and background processing?
>
>Because of what you write. You very strongly imply that a single
>application means that no task switching is neccessary.
I think it is a strong desire to interpret it that way, more than any
intent to imply that from my end. Task switching, however, is performed
in user time. There are more than enough cycles to completely change
the method of prioritizing processes when the user task switches. It is
having a single static method at all times which seems the more limited
approach, not considering a single static set of processes. I have
never thought task switching was relevant, from your perspective *or*
mine.
>> (As if I didn't know...)
>
>It does appear, on this side of the screen, that you do not.
Yes, I did know why you insist I'm hand-waving task-switching or not
considering 'background' processes. Because it seems to make sense that
I must not be aware of them, since in that context, you can most easily
avoid having to comprehend my remarks. If you want to know why it looks
like I do not know what I'm saying, look in a mirror. I'm trying to
discuss something, but I don't seem to be getting any response apart
from "that does not need to be discussed and you are obviously
misunderstanding something if you think it does." So, what the hell. I
have such a short attention span, that I have already ignored my intent
to let this drop several times. I'm already ahead of the game, having
figured out many issues so subtle it would take six months of this kind
of crap before you'd be willing to listen to them. So I give up.
>> >Ah . . . I see you prefer PMT over CMT in your automobiles . . .
>>
>> So PMT is both automatic and manual transmission?
>
>No. PMT is a manual transmission, as it allows you greater control than
>an automatic (CMT) system does.
The former analogy was that PMT was the automatic transmission; far more
complicated than the end user needs to understand, but still benefitting
them. The vast array of issues brought up by this counter-analogy will
be, I'm afraid, left unrevealed.
Bye.
--
T. Max Devlin
Manager of Research & Educational Services
Managed Services
ELTRAX Technology Services Group
[EMAIL PROTECTED]
-[Opinions expressed are my own; everyone else, including
my employer, has to pay for them, subject to
applicable licensing agreement]-
====== Posted via Newsfeeds.Com, Uncensored Usenet News ======
http://www.newsfeeds.com - The #1 Newsgroup Service in the World!
======= Over 80,000 Newsgroups = 16 Different Servers! ======
------------------------------
From: mlw <[EMAIL PROTECTED]>
Crossposted-To: comp.os.ms-windows.nt.advocacy
Subject: Re: What I've always said: Netcraft numbers of full of it
Date: Sat, 22 Jul 2000 15:33:28 -0400
Drestin Black wrote:
>
> "sandrews" <[EMAIL PROTECTED]> wrote in message
> news:[EMAIL PROTECTED]...
> > Drestin Black wrote:
> > >
> > > "Craig Kelley" <[EMAIL PROTECTED]> wrote in message
> > > news:[EMAIL PROTECTED]...
> > > > "Drestin Black" <[EMAIL PROTECTED]> writes:
> > > >
> > > > > so tell us, specifically, why servers running Windows could not do
> what
> > > the
> > > > > servers running linux @ Google do? Forget needing 6000 boxes. Why
> would
> > > the
> > > > > OS be a limitation? Impress us by showing ANY technical knowledge
> > > > > whatsoever!
> > > >
> > > > Someone sounds grumpy because Microsoft was caught cheating on their
> > > > Windows 2000 launch TPC-C score. :>
> > >
> > > caught cheating? Hahahaha - that's a twisted way of putting it. Not
> > > cheating, but an error. and the results were pulled and no one bitched.
> MS
> > > simply fixed the error in their *beta* SQL server and are as we speak
> > > retesting with Compaq. Meanwhile MS and IBM (who had to fix the
> iDENTICAL
> > > problem in their db2 software (no doubt planning on "cheating too"))
> turned
> > > around and delivered a TPC-C score twice as high as the compaq scores -
> four
> > > times better than the next *nix score at half the price!
> > >
> > > >
> > > > (not that such meters are important, mind you)
> > >
> > > oh but they do - as they did to the unix weenies before they lost and
> lost
> > > huge - now they feign lack of interest... hahahaahah
> >
> >
> > If you knew anthing about RDBMs you would know the os makes no
> > difference on
> > the outcome, THe RDBM doesn`t really use the os.
>
> HAHAHA - excuse me? I think that you are completely misinformed. You are
> wrong. Totally. That's like saying: RDBM doesn't use files or memory.
Actually, most high end DBMS systems circumvent the file system by
accessing the lowlevel block device directly. Also, they usually
allocate a huge amount of RAM and manage the memory internally.
--
Mohawk Software
Windows 9x, Windows NT, UNIX, Linux. Applications, drivers, support.
Visit http://www.mohawksoft.com
Nepotism proves the foolishness of at least two people.
------------------------------
Date: Sat, 22 Jul 2000 15:39:58 -0400
From: Gary Hallock <[EMAIL PROTECTED]>
Crossposted-To: comp.sys.mac.advocacy,comp.os.ms-windows.advocacy,comp.unix.advocacy
Subject: Re: Linsux as a desktop platform
"T. Max Devlin" wrote:
>
>
> Quite possible. If you have any suggestions, I'd be happy to hear them.
> I think it may be so contrary to your experience that you might consider
> any statement of it to be difficult to understand.
>
There you go with the insults again!
>
>
> Again, I think this resolves on a preconception of a singular task,
> which I didn't take for granted. I would suggest that it could be
> reasonable to call any system which this statement applies to a
> "workstation". On a desktop, not only is the interactive task the most
> important, it is often the only end-user task to speak of.
>
You have a very narrow view of "desktop", one that is quite antiquated.
I always have many interactive tasks running simultaneously on my desktop.
PMT allows me to switch between them instantaneously (by that I mean so fast
that the human senses can not detect the time). CMT does not allow that.
> The priority of tasks is entirely independent, and certainly not
> determined, by CPU cycles. Tasks have priority for reasons end-users
> (and their bosses, often) determine, not CPUs and engineering
> efficiencies.
Hence PMT.
Gary
------------------------------
From: T. Max Devlin <[EMAIL PROTECTED]>
Crossposted-To: comp.sys.mac.advocacy,comp.os.ms-windows.advocacy,comp.unix.advocacy
Subject: Re: Linsux as a desktop platform
Date: Sat, 22 Jul 2000 15:40:38 -0400
Reply-To: [EMAIL PROTECTED]
Said John W. Stevens in comp.os.linux.advocacy;
>"T. Max Devlin" wrote:
>>
>> Said Christopher Smith in comp.os.linux.advocacy;
>> [...]
>> >The reason he is being "an asshole" is because you simply refuse, despite
>> >numerous suggestions, to even learn the absolute and most basic differences
>> >between PMT and CMT.
>>
>> I never "refused" to learn anything in my life.
>
>Did you read my simplified explanation on how PMT works, yet?
'Yet'? I read it when you posted it, AFAIK. I think I recall it was
very helpful, thanks. Perhaps I should have said "I never refused to
learn *from* anything in my life", but that didn't reflect the asinine
comment I was responding to well enough. I do always refuse, entirely,
to "learn" anything just because somebody else said it was so, I'm
afraid. I need to understand it in some comprehensive way, before I'm
willing to accept that both the presentation and the underlying concepts
are comprehensive and complete; such is my nature. I've often remarked
that you don't really know something until you can understand why its
opposite is also true. Context, after all, is everything. Memorizing
things based on referral to authority is not "learning", in my book.
That's just making assumptions that something is right because it "seems
to make sense" in the context of previous assumptions, and I've seen far
too many examples of where that has caused problems.
Consider, for instance, if you were to base everything you know on one
book, based on the premise, which we will presume is correct, that
everything in that book is true. Does that mean you know everything?
In order to comprehend something, you must synthesize information from
multiple sources, *regardless* of how correct, comprehensive, or even
complete, that source of information is.
Did I memorize your presentation of PMT, and assume that each and every
point was absolutely true and correct? No. Did I learn from it? Of
course. I've never refused to learn anything in my life.
But I will admit that sometimes other people seem to take delight in
making it not worth the effort.
Bye.
--
T. Max Devlin
Manager of Research & Educational Services
Managed Services
ELTRAX Technology Services Group
[EMAIL PROTECTED]
-[Opinions expressed are my own; everyone else, including
my employer, has to pay for them, subject to
applicable licensing agreement]-
====== Posted via Newsfeeds.Com, Uncensored Usenet News ======
http://www.newsfeeds.com - The #1 Newsgroup Service in the World!
======= Over 80,000 Newsgroups = 16 Different Servers! ======
------------------------------
From: T. Max Devlin <[EMAIL PROTECTED]>
Crossposted-To:
comp.os.ms-windows.advocacy,comp.os.ms-windows.nt.advocacy,alt.fan.bill-gates,soc.singles,alt.fan.rush-limbaugh
Subject: Re: If Microsoft starts renting apts (was: If Microsoft starts renting apps)
Date: Sat, 22 Jul 2000 15:53:01 -0400
Reply-To: [EMAIL PROTECTED]
Said Aaron R. Kulkis in alt.fan.bill-gates;
>> All of this discussion about Microsoft renting apps with .NET
>> got me to thinking...what are we facing if, in fact, Microsoft
>> does start renting apts???
>>
>> Microsoft Apts 2000
>>
>> WINDOW
>>
>> No apartment may ever have more than one window. Residents might
>> forget which window they were looking out of and get confused.
[...]
Hilarious! Where'd it come from?
--
T. Max Devlin
-- Such is my recollection of my reconstruction
of events, as I recall. Consider it.
Research assistance gladly accepted. --
====== Posted via Newsfeeds.Com, Uncensored Usenet News ======
http://www.newsfeeds.com - The #1 Newsgroup Service in the World!
======= Over 80,000 Newsgroups = 16 Different Servers! ======
------------------------------
From: Graham Murray <[EMAIL PROTECTED]>
Crossposted-To: gnu.misc.discuss
Subject: Re: Richard Stallman's Politics (was: Linux is awesome!
Date: 22 Jul 2000 19:14:38 +0000
In gnu.misc.discuss, [EMAIL PROTECTED] (phil hunt) writes:
> What if an employee sets up on his own, as a competitor to his former
> employer, using contacts he gained in his last job?
Is that necessarily a bad thing? It may be that the contacts became
clients of the former employer because of the employee. In such
circumstances, is it wrong that the employee takes some of "his"
clients with him? One example might be a hair stylist, where many
people patronise a particular salon because they like the work of the
particular stylist.
------------------------------
** FOR YOUR REFERENCE **
The service address, to which questions about the list itself and requests
to be added to or deleted from it should be directed, is:
Internet: [EMAIL PROTECTED]
You can send mail to the entire list (and comp.os.linux.advocacy) via:
Internet: [EMAIL PROTECTED]
Linux may be obtained via one of these FTP sites:
ftp.funet.fi pub/Linux
tsx-11.mit.edu pub/linux
sunsite.unc.edu pub/Linux
End of Linux-Advocacy Digest
******************************
