Linux-Advocacy Digest #455, Volume #25 Wed, 1 Mar 00 12:13:10 EST
Contents:
JDK1.2.2 performance, Linux -vs- NT (John Harlow)
Re: Microsoft's New Motto (Angelos Karageorgiou)
Re: My Windows 2000 experience (Donn Miller)
Re: My Windows 2000 experience (Se�n � Donnchadha)
Re: what exactly is Linux advocacy? (John Sanders)
Re: JDK1.2.2 performance, Linux -vs- NT (Mark Hamstra)
Re: Microsoft's New Motto (The Ghost In The Machine)
Re: Video Conferencing on Linux ([EMAIL PROTECTED])
Re: How does the free-OS business model work? (Donovan Rebbechi)
Re: Microsoft's New Motto ("Neil")
Re: Binary compatibility: what kind of crack are they smoking? (Mario Klebsch)
Re: My Windows 2000 experience (5X3)
----------------------------------------------------------------------------
From: John Harlow <[EMAIL PROTECTED]>
Crossposted-To: comp.lang.java.advocacy
Subject: JDK1.2.2 performance, Linux -vs- NT
Date: Wed, 01 Mar 2000 14:15:44 GMT
I'm using the Sun JDK (1.2.2) in two environments. Both environments
run at the same time, on the same system (Dell Latitude, p3-450, 512
mRam.)
The primary environment is RH 6.1 (2.2.14) kernel with XIG commercial X
drivers. The second is NT4SP5 running in a virtual machine (VMWARE 2)
under linux. So it is sharing the memory, processor, X drivers, etc..
as a task under linux.
The curious thing is that the performance of java (running apps) is
significantly better in the Virtual NT session, than it is native to
linux. The obvious conclusion here is that the JDK for NT is more
efficient than the linux release.
My question is whether or not the difference is HotSpot? Does the
latest NT release have it. I know that the linux release does not.
Thanks
John
--
John R. Harlow
United Systems Inc.
email: [EMAIL PROTECTED]
Sent via Deja.com http://www.deja.com/
Before you buy.
------------------------------
From: Angelos Karageorgiou <[EMAIL PROTECTED]>
Crossposted-To: comp.os.ms-windows.nt.advocacy,comp.sys.mac.advocacy,comp.unix.advocacy
Subject: Re: Microsoft's New Motto
Date: Wed, 01 Mar 2000 16:34:10 +0200
mlw wrote:
>
> I'm not sure what you are talking about, but NT does multitask fairly
> well. One does not need to call wait within a thread for other threads
> to run. Yes, if I bump a thread up to real-time priority and just spin
> in a loop, the thread will get lots of time. If I just start a thread at
> normal priority, it will be just fine.
>
> If you really want to make a claim like this, show us some code that
> demonstrates what you mean. That way we can decide whether or not it is
> a flaw or feature.
Ok I am going to back off of this one, since the code is no longer
availlable to me.
--
/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\
Angelos Karageorgiou - CTO [EMAIL PROTECTED] Tel: +30 31 498104
------------------------------
Date: Wed, 01 Mar 2000 10:26:03 -0500
From: Donn Miller <[EMAIL PROTECTED]>
Crossposted-To: comp.os.ms-windows.nt.advocacy
Subject: Re: My Windows 2000 experience
Drestin Black wrote:
>
> "5X3" <[EMAIL PROTECTED]> wrote in message
> > Sure, ive seen X lock up (especially under MKlinux, DR2) but the
> > system was always recoverable via telnet or ssh.
>
> As they say, your milage may vary :)
>
> I've seen it only twice myself, one right after the other due to a stupid
> setup mistake. But it did. Had to hardware reset. I've read of others having
> the problem.
Well, in Linux, or any other type of unix, such as FreeBSD, you should
set resource usage limits. In FreeBSD, this is done via
/etc/login.conf, or in a file in the user's home directory called
$HOME/.login_conf. Here, you can set limits for CPU time and memory
usage. I think people really need to do this. I don't know how this
is done in Linux. (Linux guys?)
My experience with XFree86 3.3.x is that there's a memory leak in the
X server, the X libs, or one of the X apps (such as Netscape). By
setting resource limits, you can ensure that the system won't be
brought to its knees with excessive swapping. Of course, one possible
thing you may have been experiencing is that X or one of its apps was
sucking tons of CPU time. Therefore, your system might still have
been "up and running", but a lot of CPU and memory may have been
sapped from your system, giving the impression of a "hang". Resource
limitations would help here.
Sometimes, the tty-mode console driver for Linux (OR FreeBSD) can go
dead. So, X is hung, and you have now way of providing input to the
system. at the console. The only way out, usually, is to log into
your machine via a serial port, like with a nullmodem cable, for
example. Then, you can kill the offending process. But, the machine
may be creeping because of the excessive usage. In this case, it's
usually just best to reset the machine. Hey, if it's a desktop
machine, it may be faster to reset and have the machine boot up with
an fsck than it would be to kill the offending process if no resource
limits are set. Sometimes, it would take 5 minutes to type a letter
and have it show up on the screen.
OK, another bug I've seen is with Netscape's Motif libs and XFree86.
You see, XFree86 hsa backing store enabled by default, and it causes a
memory leak with Netscape's Motif libs. So, I suggest starting X with
startx -bs &
The "-bs" switch turns off backing store. In fact, it should never be
enabled by default anyways. (What was XFree86 thinking?)
Very rarely does X truly hang the machine. Usually, it's just an
extreme case of resource usage. However, X rarely hangs my machine
now, as I've been running XFree86 3.9.18. It's not only very fast,
but much more stable than the previous versions.
I have seen cases where a sound driver has crashed my FreeBSD machine
w/XFree86. Basically, I was playing an mp3 file with mpg123 in an
xterm. Then, I switched to a virtual console by doing ctrl alt f2.
Then, I attempted to switch back to X by doing alt f9, because X was
running on vty8, or the 9th virtual console. At that point, my
machine went dead with all kinds of flickering colors on my display.
It looked like maybe it was a PCI/ISA bus access issue. I think the
sound driver and the video drivers were trying to access the ISA/PCI
bus at the same time, which caused a crash somewhere on the PCI and/or
ISA bus.
- Donn
------------------------------
From: Se�n � Donnchadha <[EMAIL PROTECTED]>
Crossposted-To: comp.os.ms-windows.nt.advocacy
Subject: Re: My Windows 2000 experience
Date: Wed, 01 Mar 2000 10:37:38 -0500
petilon <[EMAIL PROTECTED]> wrote:
>>>
>>>Because pcAnywhere installs just like any other application.
>>
>>Which is how, exactly?
>
>pcAnywhere uses InstallShield. When you install pcAnywhere
>there is no indication that you are modifying the operating
>system. pcAnywhere doesn't warn you, neither does Windows.
>
Win2K doesn't let regular users install any software. Software
installation is a privilege that can only be granted by the
administrator. Beyond that, I don't see how it's Win2K's
responsibility to warn you when a driver gets installed. No OS I've
ever worked with ever did that. Certainly no version of Unix does, so
why are you holding Win2K to a higher standard?
>
>In other words, Windows 2000 allows any random application
>off the street to trash the OS without your knowledge.
>
Only if the user installing the app has sufficient privilege, and the
fact that it installs a system-level component isn't documented. But
what do you expect? If I'm root on a Unix box and I run the installer
of some "random application off the street", how do I know it won't
trash the OS? Answer: I don't. Unix is absolutely no better about this
than Win2K.
>>>
>>>What happened to the much touted "System File Protection"?
>>
>>What system files were lost?
>
>The device drivers, apparently.
>
Fine. What device drivers were lost?
>
>Windows 2000 apparently allows
>any random application off the street to install its own
>device drivers without your knowledge or consent.
>
So does Unix. Can you name an OS that doesn't let you install device
drivers even if you have sufficient privilege to do so?
>
>This means Windows 2000 can only be as stable and reliable as
>the last application you installed.
>
Of course, and the same holds true for Unix. That's why software
installation on both operating systems requires the privileges of
someone who knows what they're doing.
>>
>>Are you seriously suggesting that Win2K should never allow the
>>installation of device drivers?
>
>Of course not! All I am saying is that Win2K should not allow
>applications to install device drivers in a clandestine manner.
>
So what do you suggest, a warning? That wouldn't accomplish anything,
because installing a bad device driver is only one way to trash an OS.
You can also trash system files. That's why there are permissions. As
long as you have the ability to modify system files (or install
drivers), the system must assume that you know what you're doing with
respect to those files (or drivers). If you believe the system should
warn even the superuser when a driver is installed, then you must also
believe the system should issue warnings every time the superuser
installs or modifies a system file. And that's a ludicrous suggestion.
>
>Win2K should put the end-user in control, not some random
>application installer.
>
The end user *IS* in control. The installer runs on the user's behalf.
That's why the ability to install software isn't given to just anyone,
or, at least, it doesn't have to be.
>>
>>Again, what system files were modified?
>
>Are you saying device drivers don't qualify as "system files"?
>Aren't device drivers a critical piece of the operating system?
>
Fine. What device drivers were modified?
>>
>>By the way, no version of Unix can stop a user with sufficient
>>privilege from trashing the OS completely. Does that mean Unix
>>is fundamentally flawed?
>
>Unix culture is completely different. In Windows applications
>routinely copy their files into C:\WINDOWS. Unix applications
>don't do that.
>
So it's not a technical problem at all? Thank you.
>
>Take Oracle installation on Unix, for example. Oracle installer
>will not let you install the product if you are logged in as
>root. You have to log in as a regular user to install Oracle.
>(Windows applications do the exact opposite thing.) Of course
>Oracle installation requires some actions to be performed with
>root privileges. So what Oracle does is to create a very small
>shell script to do the minimal things that need root privileges.
>Then it asks you to run that shell script manually before
>continuing installation. If you like, you can now inspect the
>shell script and decide if you want to run it. If that script
>does stupid things like installing a device driver, you can
>abort the installation.
>
It looks like Oracle thought things through very well indeed, but what
does this have to do with the OS? How does Unix prevent some ISV less
considerate than Oracle from installing their app using an executable
installer (not a script) that requires root privileges and trashes
system files without warning? Answer: Unix doesn't prevent it at all.
------------------------------
From: John Sanders <[EMAIL PROTECTED]>
Subject: Re: what exactly is Linux advocacy?
Date: Wed, 01 Mar 2000 08:51:14 -0600
BRian wrote:
>
> I was looking through the Linux groups for ideas for a linux section for my
> site when this one caught my eye. What exactly is Linux Advocacy?
> BRian
It's a newsgroup where, if you are interested, you can find out why you
might like to run Linux on your PC. Its generality, however, results in
a lot of flame wars, misunderstandings and misconceptions.
I find that I like to read the threads where some Windows user writes
about how he couldn't get Linux to run on his system. Instead of going
to one of the help groups that would be apropos to his problem, they
whine and cry about their ineptitude. And they're not even embarrassed
to do so in front of all the readers here!
--
John W. Sanders
------------------------------
From: Mark Hamstra <[EMAIL PROTECTED]>
Crossposted-To: comp.lang.java.advocacy
Subject: Re: JDK1.2.2 performance, Linux -vs- NT
Date: 01 Mar 2000 10:26:08 -0500
John Harlow <[EMAIL PROTECTED]> writes:
> I'm using the Sun JDK (1.2.2) in two environments. Both environments
> run at the same time, on the same system (Dell Latitude, p3-450, 512
> mRam.)
>
> The primary environment is RH 6.1 (2.2.14) kernel with XIG commercial X
> drivers. The second is NT4SP5 running in a virtual machine (VMWARE 2)
> under linux. So it is sharing the memory, processor, X drivers, etc..
> as a task under linux.
>
> The curious thing is that the performance of java (running apps) is
> significantly better in the Virtual NT session, than it is native to
> linux. The obvious conclusion here is that the JDK for NT is more
> efficient than the linux release.
>
> My question is whether or not the difference is HotSpot? Does the
> latest NT release have it. I know that the linux release does not.
The Sun JDK for Linux doesn't use native threads or JIT. Get the
latest Blackdown release and the Borland/Inprise JIT for much
better performance.
--
Mark Hamstra
Bentley Systems, Inc.
------------------------------
From: [EMAIL PROTECTED] (The Ghost In The Machine)
Crossposted-To: comp.os.ms-windows.nt.advocacy,comp.sys.mac.advocacy,comp.unix.advocacy
Subject: Re: Microsoft's New Motto
Date: Wed, 01 Mar 2000 16:06:24 GMT
In comp.os.linux.advocacy, Stefan Ohlsson <[EMAIL PROTECTED]>
wrote on 1 Mar 2000 10:13:42 +0100
<[EMAIL PROTECTED]>:
>The Ghost In The Machine wrote:
>>
>>Contrast it with e.g., the Amiga, which was a fully multitasking
>>computer complete with networking capabilities, but has/had no concept
>>at all about different owners of files (although it does have read,
>>write, and execute privileges, as well as a few others). The Amiga
>>would make a very bad multiuser server. Similarly for MacOS (AFAIK)
>>and good old DOS.
>>
>The Amiga has multiuser support in the MuFS file system. Combined with
>AmiTCP (aka Genesis) and nice telnet/ftp daemons you have a nice remote
>login machine/server.
Ah, so the Amiga lives still! :-)
Interesting.
>
>/Stefan
>--
>[ Stefan Ohlsson ] � http://www.mds.mdh.se/~dal95son/ � [ ICQ# 17519554 ]
>
>Elgyn: My authorization code is E-A, T-M, E.
>/Alien: Resurrection
--
[EMAIL PROTECTED] -- insert random misquote here
------------------------------
From: [EMAIL PROTECTED]
Subject: Re: Video Conferencing on Linux
Date: Wed, 01 Mar 2000 16:22:22 GMT
In article <[EMAIL PROTECTED]>,
Houssam Owayed <[EMAIL PROTECTED]> wrote:
> Guys,
>
> Does any one know if there is a video conferencing tool for Linux. I
> am using Mandrake Linux 7.0 ( kernel 2.2.14). I need to access ils
> servers.
Depends on what sort of videoconferencing you want to do. There's a
clone of CU-SeeMe for Linux (QSeeMe, I think). There is also a package
called OpenH212, which is compatiable with NetMeeting.
If you're into multicasting, there is the UCL videoconferencing suite
(VIC - video, RAT - audio, SDR - session manager, WBD - whiteboard,
NTE - shared text editor). Some of these are extensions or re-writes of
LBL videoconferencing tools.
There are others, but either they're not particularly well-supported
(nv hasn't been updated in 7 years), or they offer nothing that these
don't.
Sent via Deja.com http://www.deja.com/
Before you buy.
------------------------------
From: [EMAIL PROTECTED] (Donovan Rebbechi)
Crossposted-To: gnu.misc.discuss
Subject: Re: How does the free-OS business model work?
Date: 1 Mar 2000 16:32:28 GMT
On 1 Mar 2000 13:56:38 GMT, Joseph T. Adams wrote:
>: On the other hand, I think there's a case to be made though for
>: government sponsorship of free software projects, on the grounds that
>: these projects develop infrastructure that benefits everyone.
>
>An alternative and in my opinion better (more just and more lawful)
>way of accomplishing the same goal would be legislation to require
>that federal and state governments use free / Open-Source software
>wherever possible, unless no appropriate free software exists or
>unless proprietary software would be provably more cost-effective in
>the long run (which is seldom the case due to vendorlock).
This would also possibly be a good idea.
>Financial support for free software developers would be nice, but
>there are inherent problems with government subsidies of *any* type in
>a constitutional republic,
I am not talking about direct government funding, more like an independent
body much like the NSF.
Obviously, public funding for *anything* is questionable as far as
capitalism goes, but in context of a constitutional republic, it
is a different story. ( For example, in a socialist republic, almost
everything is ul;timately government funded )
>Another potential pitfall here is that anything the federal government
>purchases, it also extensively regulates.
So you can have independent bodies. For example, the judiciary are
paid by the government, but operate autonamously ( which often causes
the government no end of annoyance )
>Excessive entanglement with governments could provide benefits, but it
You can use government funds without "excessive entanglement with
governments".
>felt worldwide, just as was the case with the stupid encryption export
>restrictions which caused not only inconvenience, but security
Ironically, a lot of research into strong encryption comes from government
grants, and the research in question is published in exportable journals.
--
Donovan
------------------------------
From: "Neil" <[EMAIL PROTECTED]>
Crossposted-To: comp.os.ms-windows.nt.advocacy,comp.unix.advocacy
Subject: Re: Microsoft's New Motto
Date: Wed, 1 Mar 2000 17:01:01 -0000
"Nick Manka" <[EMAIL PROTECTED]> wrote in message
news:89hme4$[EMAIL PROTECTED]...
> In article <89gvjq$859$[EMAIL PROTECTED]>,
> "Christopher Smith" <[EMAIL PROTECTED]> writes:
>
> >> NT is a multiuser OS in the same sense that DOS is an OS at all.
> >> Technically it's multiuser, but single-user assumptions are buried
> >> throughout the code, and they have caused problems in multiuser
> >> environments and they will continue to do so.
> >
> > Which of these single-user assumptions can you list ?
>
> Only one active login per machine which is completely tied to the
> GUI console's context
With vanilla NT, true enough.
NT TSE, or W2K with Terminal Services, slightly different.
> No support for multiple instances of user data and applications in the
> Registry
Can you expand on this bit?
Do you mean that different versions of say Office tend to use the same
hierarchy? 'cos some of that is version specific.
And per-user Office settings are largely stored in HKCU.
Some of the exceptions to this, are largely due to the vendors
implementations.
> A single flat event namespace in the kernel (they had to do major
> hacking on TSE to get around this, and still haven't, really)
Can you elaborate on this a little?
> More-or-less complete trust of the currently logged in user
I guess that rather depends on the rights of the currently logged-in user,
and ACLs on things that support ACLs.
> Expectations by programs that they are priviledged (more policy
> than than design, but still a factor if you actually
> want to run anything on your NT box)
Surely this is purely an app thing? An issue with the app design, or the
vendor's implementation?
> Heavy use of needlessly exclusive locks coupled with minimal COW behaviour
Again could you elaborate on this? I'm aware of some of the compromises of
necessity (well pretty much) within TSE.
> No user accessible global namespace and lots of indeterminable mappings
> of resources to namespaces within a user context
> (ie, network drives and printers)
Can you give some examples of this ie the problems?
> Reliance on the GUI code event handling -- if there is a problem on the
> machine you get a dialogue box for the currrently
> logged in user, and there are a few cases in which random
> unrelated services will hang waiting for event completion
> based on these. The infamous "16 open critical error dialogue
> boxes will completely hang a machine" being a good example.
I understand where you're coming from with some of this, but most people
implementing TSE (IME) knew well about this problems from previous NT 3.51
implementations (Winframe and the other derivatives), and already had a
workaround.
Neil
------------------------------
From: [EMAIL PROTECTED] (Mario Klebsch)
Crossposted-To: comp.os.linux.development.system,comp.os.linux.setup
Subject: Re: Binary compatibility: what kind of crack are they smoking?
Date: Wed, 1 Mar 2000 10:46:26 +0100
[EMAIL PROTECTED] (Stephen Harris) writes:
>Mario Klebsch ([EMAIL PROTECTED]) wrote:
>: When writing a Program for e.g. Linux 2.2.13, I cannot rely on
>You are writing code to a kernel version?
The kernel is the thing called linux, so 2.2.13 is a linux version.
>Sorry, the OS is bigger than
>just the kernel.
I know, bit Linux isn't. :-( That is what I am arguing the last
days. And IMHO Linux better becomes an OS.
>: Since Solairs is an operating system, specifying its version does
>: specify the version (and th4e interface) of all shared libraries,
>Ah - so you do agree with my point. RedHat 6.1 is an OS. "Linux 2.2.13"
>is a kernel. Code to a distribution, and you have the parallel environment.
>Redhat, SUSE, Debian et al are *DIFFERENT* operating systems.
But should the be *DIFFERENT* operating systems or should the better
be the same operating system?
>They have a
>hell of a lot in common and interoperability between them is surprisingly
>good, but they _are_ different.
IMHO, they have too much not in common.
73, Mario
--
Mario Klebsch [EMAIL PROTECTED]
------------------------------
From: [EMAIL PROTECTED] (5X3)
Crossposted-To: comp.os.ms-windows.nt.advocacy
Subject: Re: My Windows 2000 experience
Date: 1 Mar 2000 17:06:10 GMT
In comp.os.linux.advocacy 5X3 <[EMAIL PROTECTED]> wrote:
> In comp.os.linux.advocacy Erik Funkenbusch <[EMAIL PROTECTED]> wrote:
>> 5X3 <[EMAIL PROTECTED]> wrote in message
>> news:89hkc4$8su$[EMAIL PROTECTED]...
>>> In comp.os.linux.advocacy Drestin Black <[EMAIL PROTECTED]>
>> wrote:
>>>
>>> > SFP has NOTHING to do with the problem . NO system files are overwritten
>> or
>>> > corrupted. There is no DLL hell. SFP does NOT apply. What DOES apply is
>> that
>>> > Symantec, in version 8 had code in their video takeover that performed
>> big
>>> > no-nos in the new W2K driver model (that and some more goofs in their
>>> > symevnt libraries (yes, again)). Again, all Symantec's fault. AND, they
>> did
>>> > patch 8 and version 9 does not have these problems.
>>>
>>> A stable operating system does not allow a *driver* to break it utterly.
>>>
>>> See openbsd, openstep, VMS, MVS, and inferno/purgatory for details.
>> With the exception of inferno (which I know very little about). Bad drivers
>> can take down any of those OS's. Tell me, how exactly does the OS swap to
>> disk if the disk drivers crash?
> MVS and VMS can dynamically "swap" swap into any device it can see. (including
> memory space which has been loaned to a virtual machine You
> dont need functional disks, fantastically enough. You are probably correct
> regarding openbsd and openstep however, I did not take into consideration
> disk drivers.
I hate to respond to my own post here, but I recently was privvy to some first
hand anecdotal (well, semi, the guy I talked to was actually the guy that
DID it) evidence that OS390 is monstrously effective, having been able to run
63,000 virtual linux machines on ONE S390 at one time before he ran out of
memory.
I asked him about linux drivers actually being able to make S390 hang, to which
he responded: "nothing can make an S390 hang".
I almost believe him. He builds them on location for a living.
But then again, when we talk of S390s, we talk of an entirly different universe
of computing, one in which windows will *never* venture, and in which linux
will only exist as a handy interface.
p0ok
------------------------------
** 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
******************************
