Linux-Advocacy Digest #675, Volume #27 Fri, 14 Jul 00 15:13:07 EDT
Contents:
Re: C# is a copy of java (Gary Hallock)
Re: What I've always said: Netcraft numbers of full of it (abraxas)
Re: New Linux user & damn glad!! (The Ghost In The Machine)
Re: Tinman digest, volume 2451740 (EdWIN)
Re: Why use Linux? (TNT)
Re: Some Windows weirdnesses... (Cihl)
Re: Some Windows weirdnesses... (Cihl)
Re: New Linux user & damn glad!! (Cihl)
Re: Richard Stallman's Politics (was: Linux is awesome! (Jay Maynard)
Re: Web Browsers? (aflinsch)
Re: Richard Stallman's Politics (was: Linux is awesome! (Jay Maynard)
Re: Linsux as a desktop platform (John Sanders)
Re: Some Windows weirdnesses...
Re: New Linux user & damn glad!!
Re: Richard Stallman's Politics (was: Linux is awesome!
Re: Richard Stallman's Politics (was: Linux is awesome! (Leslie Mikesell)
Re: Linux lags behind Windows (Perry Pip)
----------------------------------------------------------------------------
Date: Fri, 14 Jul 2000 13:13:30 -0400
From: Gary Hallock <[EMAIL PROTECTED]>
Subject: Re: C# is a copy of java
The Ghost In The Machine wrote:
> I'm mildly surprised no one's ported GAS and GCC/G++ to VM/CMS.
> (Mind you, with Linux on an S/390, why bother? :-) )
Actually, I believe gcc (not sure about g++) does work on VM/CMS. But getting
it to integrate well with the CMS environment is a problem - we did look at gcc
a bit. I also thought about moving to OS/390 which does support C++. But,
as you said, once Linux for S/390 became a reality, there was no longer a
question of the path to take.
Gary
------------------------------
From: [EMAIL PROTECTED] (abraxas)
Crossposted-To: comp.os.ms-windows.nt.advocacy
Subject: Re: What I've always said: Netcraft numbers of full of it
Date: 14 Jul 2000 17:22:39 GMT
In comp.os.linux.advocacy [EMAIL PROTECTED] wrote:
> "Drestin Black" <[EMAIL PROTECTED]> writes:
>
>>so, they need 6000 linux boxes to achive zero downtime and perform text
>>searches? oh yea, you are REALLY impressing me now...
>
> Nope --- but they do a hell of a lot of cross-indexing.
>
> http://www.internetwk.com/lead/lead060100.htm
>
> Google's search algorithm requires massive computing power. Google
> weights each Web page for importance by analyzing the pattern of which
> pages link to others over all 300 million pages the search engine
> indexes. Google's process entails 500 million variables and 2 million
> terms to index every month, resulting in about 1 terabyte of data to
> index.
>
> The whole article is worth reading.
>
Dresden will swear up and down that windows can do it too (and will of course
be lying, as usual).
But again, as usual, his argument will collapse entirely under the burden
of proof.
=====yttrx
------------------------------
From: [EMAIL PROTECTED] (The Ghost In The Machine)
Subject: Re: New Linux user & damn glad!!
Date: Fri, 14 Jul 2000 17:24:47 GMT
In comp.os.linux.advocacy, richard harlos <[EMAIL PROTECTED]>
wrote on Fri, 14 Jul 2000 04:37:22 GMT <[EMAIL PROTECTED]>:
>Hi, all.
>
>I'm just about messin' my shorts for joy!
>
>I installed Slackware 7.1 (BigSlack, the UMSDOS install) on my PC and am
>now happily up and running on the 'net.
>
>Aside from a little tweaking to get my cheap, ISP-supplied network card
>enabled, I'm good to go.
>
>And even though it's going to take some time to learn my way around X
>and Linux in general, I'm much happier to be *doing* something about
>my dissatisfaction with Microsoft product (by not using them anymore
>than necessary!) than just *talking* about it.
>
>Don't flame this newbie too bad :)
>
>richard harlos
Flame? Nay, we salute thee, new user! :-)
Welcome to an interesting, reliable, and hopefully exciting world.
While not as highly polished as Microsoft's so-called wizards,
it may be far more interesting in some respects. I for instance
am experimenting with GTK, and, despite its flaws, is very easy
to work with, as it comes complete with code examples and bends
over backwards to ensure portability (for starters, it's written
in C). I wouldn't be surprised if there's a Windows implementation
for gtk, although I haven't looked; this means, of course, that
one could code on Linux and port to Windows, which means that
both markets are covered.
Similar issues abound in other environments; Motif, while getting
a bit on the old and hoary side, is still a strong environment
for widget development available world-wide for almost all
Unix platforms. Note also that there's freeware available;
http://www.lesstif.org
for details thereon.
I've already enabled access to our Oracle database, using nothing
but a thinclient JDBC Java driver package, freely available. This
promises an interesting future! Note also that I can prototype
at home using Postgres, and, with minimal changes, deploy
at work. This is in the future, of course -- but it does open
a door or two.
I'm also given to understand that Linux has the capability of
better firewalling than Windows 98. I could be wrong.
There's also more to protect, mind you -- Postgres in
particular has the capability of hosting itself to anyone
on the Internet, if anyone cares to set up such a capability.
Sendmail is a tad quirky. Httpd (Apache) is highly capable,
but can lead a hacker in if one is not careful regarding
security. Ftp is a potential problem and tftp is a very very
large hole waiting to be entered unless the sysadmin is very
careful.
Best put all of these behind a firewall (my firewall
machine's an old Pentium 90 with 16 megs) if you have one available,
and use IP masquerading. Or use tcp wrappers (/usr/sbin/tcpd)
and put ALL:ALL in /etc/hosts.deny, and edit /etc/hosts.allow
to suit your tastes. Edit /etc/inetd.conf if need be, to disable
them on the firewall machine.
And there's a lot of documentation to wade through. A mixed blessing,
in some respects, especially considering that it's in at least
3 places: 'man', 'info', and /usr/doc. But at least it's there.
But all these are capabilities Windows can only dream about. :-)
At least, not without extra outlay, or being employed at Redmond. :-)
--
[EMAIL PROTECTED] -- insert random misquote here
------------------------------
Subject: Re: Tinman digest, volume 2451740
From: EdWIN <[EMAIL PROTECTED]>
Crossposted-To:
comp.sys.mac.advocacy,comp.os.os2.advocacy,comp.os.ms-windows.nt.advocacy
Date: Fri, 14 Jul 2000 10:28:12 -0700
The Thole Tholenated:
>Ray Chason writes:
>
>>> Here's today's Tinman digest:
>>>
>>> 1> Incorrect.
>>>
>>> Balderdash, given your failure to comprehend the evidence
>>> presented.
>>>
>>> 1> You're merely demonstrating your difficulty in presenting
any
>>> 1> evidence.
>>>
>>> Incorrect, given that I presented it multiple times.
>>>
>>> 1> On the contrary, my answer was quite appropriate.
>>>
>>> Illogical, given that you didn't provide an answer,
therefore one
>>> cannot assess whether it was appropriate or not.
>
>> And this is on topic for any of the above groups because....
>
>Ask Tinman. I didn't choose the newsgroups.
Question not the Tholy One! Welcome the blessings of his
Tholing.
===========================================================
Got questions? Get answers over the phone at Keen.com.
Up to 100 minutes free!
http://www.keen.com
------------------------------
Subject: Re: Why use Linux?
From: [EMAIL PROTECTED] (TNT)
Date: Fri, 14 Jul 2000 17:34:35 GMT
On Fri, 14 Jul 2000 11:41:59 GMT, [EMAIL PROTECTED] (Pete Goodwin) wrote
in <8kmua5$6fe$[EMAIL PROTECTED]>:
>In article <8kmd8i$eg1$[EMAIL PROTECTED]>,
> [EMAIL PROTECTED] wrote:
>
>> Oh, I have no idea whether there is an RPM or not.
>
>There isn't, I looked.
There are Cooker RPMS of 4.0.1 compiled for Mandrake here:
ftp://rpmfind.net/linux/MandrakeCooker/contrib/RPMS/
Been there for a week already. Also check alt.os.linux.mandrake for any
previous experiences.
>
>> However, I wonder how you can use that XF86 4.0.1 install as an
>indication
>> of linux "lacking" compared to Windows --- when in fact an equivalent
>install
>> wouldn't even be possible under Windows.
>
>There's a driver for Windows 95/98 and there's a beta test one for
>Windows 2000. So it is possible.
What driver and beta test one that can change the GDI of Win9x/2000?
>
>> I don't think replacing the GDI under Windows is something you'd do
>with
>> anything but a nice, packaged patch directly from Microsoft. And yet,
>you
>> download God knows what from God knows where, install it over
>Mandrake,
>> and then complain that things don't work anymore.
>
>I got the kit from xfree86.org. Isn't that the right place?
You could have done anything wrong with the installation and configuration
for the main or optional components. You could say it's difficult, but it may
actually have to do with the way you think under Windows, the way you're too
accustomed to and dependent on the GUI to walk you arround. You could say
that's lacking, but actually installing updates and patches for MS software
almost guarantees of breaking older versions of their own, let alone third
parties', and there's nothing you can do about it but hold your breath. And I
don't see the equivalent of that under GPL/Linux softwares.
Regards,
------------------------------
From: Cihl <[EMAIL PROTECTED]>
Subject: Re: Some Windows weirdnesses...
Date: Fri, 14 Jul 2000 17:35:14 GMT
I'll give one of my annoying experiences in using Windows 98:
Logging in:
I sometimes have the tendency to make typos in my password when i log
in.
When i make such a typo in Linux, it denies my login attempt. Sounds
logical, doesn't it?
Windows' turn. Mistype the password, and Windows will ask you to
confirm your 'new' password. No way of going back to correct the
error, and if you click 'cancel' you get logged in anyway.
(Anonymously?)
-- =
=A8I live!=A8
=A8I hunger!=A8
=A8Run, coward!=A8
-- The Sinistar
------------------------------
From: Cihl <[EMAIL PROTECTED]>
Subject: Re: Some Windows weirdnesses...
Date: Fri, 14 Jul 2000 17:36:37 GMT
Stuart Fox wrote:
> =
> "V'rgo Vardja" <[EMAIL PROTECTED]> wrote in message
> news:8kmlmh$1bh2$[EMAIL PROTECTED]...
> =
> <snip everything>
> =
> Use Windows 2000 Professional...
> =
> Most Winvocates would agree with you, Win9x is a stinking pile of dog s=
hit.
> =
> Stu
Did you actually *buy* Windows 2000 Professional? Hm?
-- =
=A8I live!=A8
=A8I hunger!=A8
=A8Run, coward!=A8
-- The Sinistar
------------------------------
From: Cihl <[EMAIL PROTECTED]>
Subject: Re: New Linux user & damn glad!!
Date: Fri, 14 Jul 2000 17:41:22 GMT
Be welcome, Richard.
If you have any questions, feel free to ask. I do recommend you ask
within appropriate newsgroups, as you will get answers more quickly
and to the point.
Also, this newsgroup is here for the sole purpose of bad advocacy
(read: bashing Windows). For some good entertainment, join in anytime.
:)
-- =
=A8I live!=A8
=A8I hunger!=A8
=A8Run, coward!=A8
-- The Sinistar
------------------------------
From: [EMAIL PROTECTED] (Jay Maynard)
Crossposted-To: gnu.misc.discuss
Subject: Re: Richard Stallman's Politics (was: Linux is awesome!
Date: 14 Jul 2000 17:45:09 GMT
Reply-To: [EMAIL PROTECTED]
On Fri, 14 Jul 2000 02:23:34 GMT, Mike Stump <[EMAIL PROTECTED]> wrote:
>Another free software package, mpeg_play by Berkeley. I suspect there
>are plenty of proprietary players based upon it, and it is lagging and
>not worthwhile anymore.
If "free software" as you choose to redefine it is that important to you,
then you should pick it up and enhance it. Let your version compete in the
marketplace with the non-free ones. See which one wins.
>If you don't fear it, then, there isn't a reason to use the GPL, IMHO.
Far from fearing it, I welcome and embrace it. There's enough room for
everyone.
------------------------------
From: aflinsch <[EMAIL PROTECTED]>
Subject: Re: Web Browsers?
Date: Fri, 14 Jul 2000 13:35:33 -0500
Bob Hauck wrote:
>
> Turn off java in Netscape and see if that helps. Also, 4.73 and 4.06
> have been pretty reliable for me (even with java).
Especially if you use navigator only and not the whole communicator
package. It seems to me that the other crap added to netscape is what
brings it down sometimes.
------------------------------
From: [EMAIL PROTECTED] (Jay Maynard)
Crossposted-To: gnu.misc.discuss
Subject: Re: Richard Stallman's Politics (was: Linux is awesome!
Date: 14 Jul 2000 18:02:51 GMT
Reply-To: [EMAIL PROTECTED]
On Fri, 14 Jul 2000 04:34:17 GMT, Mike Stump <[EMAIL PROTECTED]> wrote:
>A BSDL developer is free to use GPLed software in his programs. What
>he is not free to do is to conspire to coopt the freedom that the GPL
>brings with it.
No. He can only use GPVed software in his programs if he then agrees to
license the whole of his work under the GPV. The GPV has thus infected his
work, taking it out of his control. This is hardly freedom; it's coercive
re-licensing of someone else's work, and the exact thing I've been
complaining about all along.
You and the other GPVists here have been proclaiming that people should be
fre to adopt the GPV for their own work, yet you continually refuse to
acknowledge that others have reasons for not doing so, and you continually
refuse to allow them the same right.
------------------------------
From: John Sanders <[EMAIL PROTECTED]>
Crossposted-To: comp.sys.mac.advocacy,comp.os.ms-windows.advocacy,comp.unix.advocacy
Subject: Re: Linsux as a desktop platform
Date: Fri, 14 Jul 2000 13:06:56 -0500
"T. Max Devlin" wrote:
[deletia]
> Tell me, because I'm not really that familiar with this bit, how would a
> user that wanted to tell their computer "I want this to speed up" do it,
> and how would the computer know when that was no longer necessary. How
> would a user say "I want this app to have a higher priority every time I
> start it." If these are easily understood and manipulated controls,
> than half of my argument disappears.
man rtprio.
[deletia]
> --
> 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]-
--
John W. Sanders
===============
"there" in or at a place.
"their" of or relating to them.
"they're" contraction of 'they are'.
------------------------------
From: <[EMAIL PROTECTED]>
Subject: Re: Some Windows weirdnesses...
Date: Fri, 14 Jul 2000 09:36:28 -0700
Reply-To: <[EMAIL PROTECTED]>
The wierdest thing about all the security problems that you cite is that
they were all solved years ago by unix on the one hadn and Novell's NetWare
on the other. Novell was Microsoft's biggest competitor on the area of
networking PC. In the days of Dos and Windows 3.x workstations working off
of a Novell NetWare fileserver, none of those problems were possible. When
Microsoft bundled netoworking into Windows they had the those other
pre-existing system to show them what they needed to do to make Microsoft
Networking secure and safe. Every networking platform became more secure
and flexible than the previous generations. At that time Lan Manager was
the joke of the industry, when they incorporated Lan Manager into Windows
and called it Microsoft Networking, that was their opportunity to fix it,
but instead they broke it even more.
Now let me contribute a couple of more items to your list of Windows
weirdnesses.
On a workstation running Windows 95 in which the task bar is set to
autohide, has popped up from the bottom of the screen for years. This year
it started popping down from the top of the screen instead. When it changed
its position there are no system alterations going on, the only thing
running was an ordinady user applicaton. Nothing could make it return to
its original position. Including restoration of the registery and ini's to
a prior state from backup. Then just a couple of days ago, after months it
return to normal just an mysteriously.
On a multi boot workstation, dos, BSD, Linux, and others boot just fine and
work flawlessly. Windows 95 will generate windows protection error about
half the time when it is warm booted and everytime it is cold booted. For
three months this changed and Windows 95 booted just fine every time. Then
it started to WPE again when booting. There have been no hardware
modifications and not installation or reconfiguration of Windows software on
that workstation from before Window's change of behavior when booting.
------------------------------
From: <[EMAIL PROTECTED]>
Subject: Re: New Linux user & damn glad!!
Date: Fri, 14 Jul 2000 10:58:48 -0700
Reply-To: <[EMAIL PROTECTED]>
NE2000's have never given me any trouble with Linux or anything. Only
Windows has had some fits with different versions of it.
But a your comment of a latter version of Linux working when a prior one did
not, reminds me of my first couple of attempts at installing Linux as well.
That was in the days of Slackware 1.0. I purchased a computer book, not a
Linux book (they did not exist yet). The book came with a CD that had data
the supported the purpose of the book. They had some room left over on the
CD so they put a copy of Slackware 1.0 on it as well. I didn't discover
this until after I got the CD into my drive. I was pleased since I had
heard about Linux and was interested in getting my hand onto a unix host
again.
Once I examined the system requirement and supported hardware and
installation procedures I was all ready to give it a try. but there was
one problem, my harddrive at the time was SCSI and Linux did support my SCSI
host adaptor, which was a very popular model at the time. Even better, one
of the boot disks had support for the SCSI host adaptor compiled into the
kernel; but the publisher of the book didn't include that boot disk image on
the CD. I was out of luck.
If it were today, it would have been easy to just download the correct boot
floppy image file and everything would have been alright. But not in those
days, unless you were lucky, a SLIP or PPP access to the net used to start
from aroung $500.00/month for individuals and decent UUCP, or shell account
also were not quite as bad but they were costly as well. The other solution
of the time was calling a local BBS that could route mail through a fidonet
to internet gateway or had a UUCP connection. I could then get the image
file be sending private mail to the gateway which would be translated into
UUCP or SMTP email directed to a FTPMAIL server which would contact an
archie server to locate the file and mail me the results. Then send more
mail to the FTPMAIL server to direct it to get the files UUENCODE it and
mail it to me in 100 line segments that woud each me mailed as sepperate
emails so that it could transverse the gateways and networks. Which I would
then assemble and decode. But those gateways and FTPMAIL server were very
busy and it could take months to get your files, unless you were lucky. If
a segment or two of the file were lost or scrambled you would have to
request it again and wait some more.
Before I got this missing file by the above method, I purchased another book
that had a CD and I found another copy of Slackware 1.0 on it. This copy
also had missing parts, but the one file I needed was on it. Linux through
the Slackware 1.0 distribution installed just fine.
Two week latter I got the last segment of the file by mail from the BBS.
------------------------------
From: [EMAIL PROTECTED] ()
Crossposted-To: gnu.misc.discuss
Subject: Re: Richard Stallman's Politics (was: Linux is awesome!
Date: Fri, 14 Jul 2000 18:35:39 GMT
On 14 Jul 2000 18:02:51 GMT, Jay Maynard <[EMAIL PROTECTED]> wrote:
>On Fri, 14 Jul 2000 04:34:17 GMT, Mike Stump <[EMAIL PROTECTED]> wrote:
>>A BSDL developer is free to use GPLed software in his programs. What
>>he is not free to do is to conspire to coopt the freedom that the GPL
>>brings with it.
>
>No. He can only use GPVed software in his programs if he then agrees to
>license the whole of his work under the GPV. The GPV has thus infected his
>work, taking it out of his control. This is hardly freedom; it's coercive
It depends entirely upon who you ask.
[deletia]
Many of the rest of us don't have any problem with people who
would seek to use common code as if it were their own personal
property (with all that implies in software) being restricted.
--
The LGPL does infact tend to be used instead of the GPL in instances
where merely reusing a component, while not actually altering that
component, would be unecessarily burdensome to people seeking to build
their own works.
This dramatically alters the nature and usefulness of Free Software
in practice, contrary to the 'all viral all the time' fantasy the
anti-GPL cabal here would prefer one to believe.
|||
/ | \
------------------------------
From: [EMAIL PROTECTED] (Leslie Mikesell)
Crossposted-To: gnu.misc.discuss
Subject: Re: Richard Stallman's Politics (was: Linux is awesome!
Date: 14 Jul 2000 13:47:16 -0500
In article <[EMAIL PROTECTED]>,
T. Max Devlin <[EMAIL PROTECTED]> wrote:
> [...]
>>Please stop revealing your ignorance; as someone who *writes* software
>>for a living and deals with the improvement of 'a product that's
>>already written' every single fucking day ... I know what software
>>costs to write, and it ain't what you pretend to think. (I'm not sure I
>>give you the benefit of the doubt on thinking at this point. You seem
>>to have bought your own hype.)
>
>I know what software costs to right, too, and I'm not even a developer.
>It costs about 1/100, at best, of what most companies sell it for.
If anyone could sit down and write the best-selling program without
having to discard the other hundred attempts that most places
go through, he would deserve to be rich. Generally that's
not the way it happens.
Les Mikesell
[EMAIL PROTECTED]
------------------------------
From: [EMAIL PROTECTED] (Perry Pip)
Subject: Re: Linux lags behind Windows
Reply-To: [EMAIL PROTECTED]
Date: Fri, 14 Jul 2000 18:41:30 GMT
On Thu, 13 Jul 2000 19:31:16 -0400,
Aaron R. Kulkis <[EMAIL PROTECTED]> wrote:
>
>
>Pete Goodwin wrote:
>>
>> In article <[EMAIL PROTECTED]>,
>> "Aaron R. Kulkis" <[EMAIL PROTECTED]> wrote:
>>
>> If you're talking multiprocess or multithreaded processing then I can
>> see why race conditions are important, but I still don't see why anyone
>> would need to know how the scheduler works.
>>
>>
>> Previously you said I would need to know about the scheduler. Now you
>> seem to think I need to know about race conditions.
>>
>> Please expand on what you mean. I can't see that every application
>> needing to consider race conditions - which is what you appear to be
>> saying.
>
>I never implied that *EVERY* application has race conditions.
>However, any programmer who keeps himself IGNORANT of process
>scheduling and race conditions is going to step on a land mine,
>and NEVER figure it out.
>
There is a difference between knowing what the scheduler might do and
the knowing internals of how the scheduler works. For race conditions
you need to know what the scheduler might or might not do. For hard
real-time applications you need to know and possibly control
deterministicly what the scheduler does. For soft real time apps, or
in cases where performance is an issue, it's a big help to understand
the scheduler more deeply.
>
>Multi-process simulations.
I am currently writing code for hard real-time high-fidelity flight
simulations that will be used in a number of engineering
testbeds. These include testbeds for embedded flight software
verification, flight dynamics verification, avionics hardware
integration, and crew training. It's part of the design and
development of an experimental research vehicle scheduled to
fly in 2002.
>Let's suppose I'm simulating a battlefield, and, to make the task of
>modelling each entity, I have decided to model each unit, weapon,
>building, even the terrain, as a seperate processes.
Uhm...The first question that comes to my mind is does this sim need
to be real time?? If it's just a pure simulation for analysis,
probably not at all. If it's an action game, probably soft real
time. If it's integration testing for some new Army "smart weapon", it
may need to run in hard real time. Just how much real time performance
you need determines how much you really need to know/control the
scheduling.
<warning - alternate OS advocacy rant> For hard real performance we
use IRIX's Real time extensions from SGI:
http://www.sgi.com/Products/React.html which provides to the user full
control of of up to all but one system processor for dedication to
real time tasks. (Leaving at least processor left behind for the OS
and daemons). You can also redirect hardware interrupts towards or
away from your realtime processors, depending on whether they are from
real-time critical devices or from devices you don't want to
interference from. Thus you can take complete control of both
scheduling and user level interrupt handling on those processors.
And IRIX supports up to 512 CPU's. So that's up to 511 processors
available for real time tasks. </warning - alternate OS advocacy
rant>
>Now...all you have to do is to have each process write to an "input"
>pipe or socket of another process which it is interacting with.
Well for even soft real time I would recommend shared memory and
semaphores. On distributed realtime testbeds we use reflective memory
rings, which have latencies as low as a few microseconds, hundreds of
times faster than sockets/ethernet.
>Let's say that all messages to the map consist of 3 parts
>
>A) Header (which Process is sending this message?)
>B) Map Coordinate
>C) Information request (elevation, terrain type, etc.)
> or Action (small-arms fire, tank running through building, etc.)
>
>
>Let's say we have 3 processes:
OK
>Process A which is a "target" for the actions of two different
>processes.
>
>Process 100 is the map, and listens to pipe /tmp/pipe_100
>Process 105 is a unit which is moving, and needs to ask the map
>what kind of terrain it is traversing.
>Process 123 is another unit which is firing a cannon at some
>some point on the map.
>
>Process 123 writes the "preamble" of it's message to /tmp/pipe_100.
>Before it gets to the main body of it's message, Process 123 loses
>the CPU. In the intervening time, Process 105 writes to /tmp/pipe_100
>with a simple request for terrain information at it's location.
>Sometime later, Process 123 finishes writing it's message to
>/tmp/pipe_100
>
You don't even have to lose the CPU to get race condition on a
multiprocessor system.
>[race condition for dummies snipped]
>
>Someone who doesn't understand the task scheduler is likely to
>allow this sort of situation.
Well certainly not to much depth of knowledge of the scheduler to
prevent race conditions. All the devloper needs to know about
scheduling it that he can't predict what the scheduler will do.
>Conversely, someone who DOES understand process scheduling will
>realize that
>
>a) write(2) is guaranteed to NOT be interrrupted by a context switch.
> that is...the context switch will be ignored until the write(2)
> requests transfers the data into the filesystem I/O buffers.
NOT!! There is a maximum number of bytes that you can write(2) to a
pipe or FIFO and still be guaranteed atomicity. Above that number,
called PIPE_BUF as required by Posix, their is no guarantee you will
get a full write. In that case, write(2) will return the number of
bytes written. PIPE_BUF is defined in limits.h in your Linux kernel
source. So now there are some unseen bugs in your code if your
messages are too long.
>b) printf(3) is NOT guaranteed to transfer the output into the
> buffers completely uninterrupted by context switches.
I think you mean fprintf(3), printf(3) writes to stdout.
>c) the use of printf(3) allows the race condition above to occur.
>
>d) the use of 3 seperate writes (proc ID, co-ordinate n-tuple,
> and request/event) ALSO allows the race condition above
> to occur
>
and e) if your messages are larger than PIPE_BUF, allows the race
condition above to occur.
>therefore
>
>e) EACH PROCESS ***MUST*** scribble it's own output into it's
> own internal buffer,
and verify the message is smaller than or equal to PIPE_BUF.
and then ***MUST*** use write(2)
> to transfer the entire message in ONE PIECE to the
> filesystem pipe.
>
>The programmer who is ignorant of the operation of the scheduler
>is going to have unexplained errors which he will NEVER NEVER
>NEVER understand until he learns about process scheduling.
And you who is ignorant of limits will have unexplained errors.
>
>ANOTHER reason to learn the process scheduler's algorithm is to
>prevent processes from suffering from CPU starvation.
>
Don't need to know that much about the scheduler, unless the app is
real time. Do need to know about limits.
>Aaron R. Kulkis
>Unix Systems Engineer
Engineer?? As in a four year degree that ends in the Letter E, for
Engineering?? I've got a BSME and am working on a MSCS.
Perry
------------------------------
** 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
******************************
