WinModems
Matthias Blankertz
modems anywhere on the internet?
Godwin Stewart
Hash: SHA1
Matthias Blankertz wrote:
> is there a list of which modems are winmodems and which models are real
> modems anywhere on the internet?
Possibly.
However, there are very few exceptions to the rule which says "internal PCI
boards and external USB devices are WinModems, internal ISA boards and
external serial devices are Modems".
- --
G. Stewart -- gstewart at gstewart dot homeunix dot net
gstewart at spamcop dot net
Registered Linux user #284683
GnuPG key : BA3D01C6 (pgp.mit.edu)
Fingerprint: C3DF C686 6572 6E59 E3E4 0F40 2B9A 2218 BA3D 01C6
- ---------------------------------------------------------------
Don't be irreplaceable. If you can't be replaced, you
can't be promoted.
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.1 (GNU/Linux)
iD8DBQE+K/uSK5oiGLo9AcYRAuYOAJ9x3O84y7B9TARHhQb5kjxmkcItDwCggrcl
hII36KHHxR2lpF0ffGvHl08=
=8DNk
-----END PGP SIGNATURE-----
mps
> is there a list of which modems are winmodems and which models
> are real modems anywhere on the internet?
google for winmodems...
linmodems.org has a huge database
Jean-David Beyer
> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
>
> Matthias Blankertz wrote:
>
>
>>is there a list of which modems are winmodems and which models are real
>>modems anywhere on the internet?
>
>
> Possibly.
>
> However, there are very few exceptions to the rule which says "internal PCI
> boards and external USB devices are WinModems, internal ISA boards and
> external serial devices are Modems".
That is very disappointing. Were I to get a PCI-only machine, would I be
doomed to getting a win-modem? I cannot see why anyone would bother
buying a machine with a PCI backplane and wanting a WinModem. Of course,
I cannot understand why anyone would want a brain-damaged modem anyway.
Perhaps they were designed to use up spare cycles for people running web
browsers only on 1200MHz CPU systems.
>
> - --
> G. Stewart -- gstewart at gstewart dot homeunix dot net
> gstewart at spamcop dot net
> Registered Linux user #284683
>
> GnuPG key : BA3D01C6 (pgp.mit.edu)
> Fingerprint: C3DF C686 6572 6E59 E3E4 0F40 2B9A 2218 BA3D 01C6
> - ---------------------------------------------------------------
> Don't be irreplaceable. If you can't be replaced, you
> can't be promoted.
> -----BEGIN PGP SIGNATURE-----
> Version: GnuPG v1.2.1 (GNU/Linux)
>
> iD8DBQE+K/uSK5oiGLo9AcYRAuYOAJ9x3O84y7B9TARHhQb5kjxmkcItDwCggrcl
> hII36KHHxR2lpF0ffGvHl08=
> =8DNk
> -----END PGP SIGNATURE-----
--
.~. Jean-David Beyer Registered Linux User 85642.
/V\ Registered Machine 73926.
/( )\ Shrewsbury, New Jersey http://counter.li.org
^^-^^ 9:10am up 7 days, 17:18, 2 users, load average: 2.24, 2.27, 2.21
Peter T. Breuer
> That is very disappointing. Were I to get a PCI-only machine, would I be
> doomed to getting a win-modem? I cannot see why anyone would bother
No, there are some (very few) pci hardware modems. You'll have to
google for them - I forget the brand name.
Peter
Godwin Stewart
Hash: SHA1
Jean-David Beyer wrote:
> That is very disappointing. Were I to get a PCI-only machine, would I be
> doomed to getting a win-modem?
Not necessarily. I didn't say that there were *no* exceptions to the rule,
just that they are few and far between.
In any event, when searching around for a PCI modem board, _NEVER_ listen to
what the vendor has to say, research it for yourself on the web. I've seen
vendors tell people that USB modems are faster than serial modems just to
pitch a sale, so I can imagine them also saying that PCI modems are better
than ISA modems because they're on a faster bus.
> I cannot see why anyone would bother buying a machine with a PCI backplane
> and wanting a WinModem.
They might not have any choice. There are fewer and fewer mobos around with
ISA slots.
> Of course, I cannot understand why anyone would want a brain-damaged modem
> anyway.
"Brain-damaged"? What brain? :)
> Perhaps they were designed to use up spare cycles for people running web
> browsers only on 1200MHz CPU systems.
LOL :)
- --
G. Stewart -- gstewart at gstewart dot homeunix dot net
gstewart at spamcop dot net
Registered Linux user #284683
GnuPG key : BA3D01C6 (pgp.mit.edu)
Fingerprint: C3DF C686 6572 6E59 E3E4 0F40 2B9A 2218 BA3D 01C6
- ---------------------------------------------------------------
Do molecular biologists wear designer genes?
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.1 (GNU/Linux)
iD4DBQE+LBYuK5oiGLo9AcYRAgMdAJ0SZ0etbQjIObBdbfGcf2MsfnuVigCYmHJJ
kKXeF9cgP93dKJnYJGWIGA==
=1SXi
-----END PGP SIGNATURE-----
John Hasler
> No, there are some (very few) pci hardware modems. You'll have to google
> for them - I forget the brand name.
I believe Multitech makes one.
--
John Hasler
jo...@dhh.gt.org
Dancing Horse Hill
Elmwood, Wisconsin
Robert Heller
In a message on Mon, 20 Jan 2003 15:27:55 +0100, wrote :
"TB> Jean-David Beyer <jdb...@exit109.com> wrote:
"TB> > That is very disappointing. Were I to get a PCI-only machine, would I be
"TB> > doomed to getting a win-modem? I cannot see why anyone would bother
"TB>
"TB> No, there are some (very few) pci hardware modems. You'll have to
"TB> google for them - I forget the brand name.
And they tend to be more costly than an *external* RS232 modem.
"TB>
"TB> Peter
"TB>
Robert Heller
In a message on Mon, 20 Jan 2003 09:16:24 -0500, wrote :
JB> Godwin Stewart wrote:
JB> > -----BEGIN PGP SIGNED MESSAGE-----
JB> > Hash: SHA1
JB> >
JB> > Matthias Blankertz wrote:
JB> >
JB> >
JB> >>is there a list of which modems are winmodems and which models are real
JB> >>modems anywhere on the internet?
JB> >
JB> >
JB> > Possibly.
JB> >
JB> > However, there are very few exceptions to the rule which says "internal PCI
JB> > boards and external USB devices are WinModems, internal ISA boards and
JB> > external serial devices are Modems".
JB>
JB> That is very disappointing. Were I to get a PCI-only machine, would I be
JB> doomed to getting a win-modem? I cannot see why anyone would bother
JB> buying a machine with a PCI backplane and wanting a WinModem. Of course,
JB> I cannot understand why anyone would want a brain-damaged modem anyway.
JB> Perhaps they were designed to use up spare cycles for people running web
JB> browsers only on 1200MHz CPU systems.
All External, RS232 (serial port), Hayes AT command set modems are 100%
compatible with Linux. All you need is a *standard* serial port. Even
if your computer lacks a serial port there are suported PCI serial
cards.
WinModems were designed to save one socket on your power strip, save one
cable, to use of spare cycles on machines processors running at >
200mhz. *AND* to shave about US$10 off the cost of production (US$10 ==
roughly the cost of a embedded microprocessor w/ UART).
JB>
JB> >
JB> > - --
JB> > G. Stewart -- gstewart at gstewart dot homeunix dot net
JB> > gstewart at spamcop dot net
JB> > Registered Linux user #284683
JB> >
JB> > GnuPG key : BA3D01C6 (pgp.mit.edu)
JB> > Fingerprint: C3DF C686 6572 6E59 E3E4 0F40 2B9A 2218 BA3D 01C6
JB> > - ---------------------------------------------------------------
JB> > Don't be irreplaceable. If you can't be replaced, you
JB> > can't be promoted.
JB> > -----BEGIN PGP SIGNATURE-----
JB> > Version: GnuPG v1.2.1 (GNU/Linux)
JB> >
JB> > iD8DBQE+K/uSK5oiGLo9AcYRAuYOAJ9x3O84y7B9TARHhQb5kjxmkcItDwCggrcl
JB> > hII36KHHxR2lpF0ffGvHl08=
JB> > =8DNk
JB> > -----END PGP SIGNATURE-----
JB>
JB>
JB>
JB> --
JB> .~. Jean-David Beyer Registered Linux User 85642.
JB> /V\ Registered Machine 73926.
JB> /( )\ Shrewsbury, New Jersey http://counter.li.org
JB> ^^-^^ 9:10am up 7 days, 17:18, 2 users, load average: 2.24, 2.27, 2.21
JB>
JB>
Johan Kullstam
> Godwin Stewart wrote:
> > -----BEGIN PGP SIGNED MESSAGE-----
> > Hash: SHA1
> > Matthias Blankertz wrote:
> >
> >>is there a list of which modems are winmodems and which models are real
> >>modems anywhere on the internet?
> > Possibly.
> > However, there are very few exceptions to the rule which says
> > "internal PCI boards and external USB devices are WinModems,
> > internal ISA boards and external serial devices are Modems".
>
> That is very disappointing. Were I to get a PCI-only machine, would I
> be doomed to getting a win-modem? I cannot see why anyone would bother
> buying a machine with a PCI backplane and wanting a WinModem. Of
> course, I cannot understand why anyone would want a brain-damaged
> modem anyway. Perhaps they were designed to use up spare cycles for
> people running web browsers only on 1200MHz CPU systems.
I think computers are still shipping with RS-232 serial ports. Go
external.
--
Johan KULLSTAM <kulls...@attbi.com> sysengr
#Harold Stevens US.972.952.3293
[Snip...]
>And they tend to be more costly than an *external* RS232 modem.
And they tend to take down everything else for a power cycle (while the
external can be turned off independently, if needed).
(I have an ISA internal at home which will not detect carrier every few
months or so until it--and thus the entire box--is power cycled).
--
Regards, Weird (Harold Stevens) * IMPORTANT EMAIL INFO FOLLOWS *
Pardon any bogus email addresses (mklog*) in place for spambots.
Really it's (wyrd) at raytheon, dotted with com. DO NOT SPAM IT.
Standard Disclaimer: These are my opinions not Raytheon Company.
John Hasler
> I have an ISA internal at home which will not detect carrier every few
> months or so until it--and thus the entire box--is power cycled
I had one of those. I just soldered in an external reset switch.
There is really no excuse for firmware of that sort hanging, though.
#Harold Stevens US.972.952.3293
[Snip...]
>There is really no excuse for firmware of that sort hanging, though.
I agree completely, and I've seen symptoms like this on only one other ISA
modem in several years, IIRC.
BTW, I tried all kindsa "AT" reset combos I found while running it down (I
don't recall exactly the commands, and apparently have no records now) but
to no avail. On occasion, these suckers just seem to need a poweroff reset
(AFAICT).
Tips and suggestions for troubleshooting this quirk welcome.
Peter T. Breuer
> In <87iswjo...@toncho.dhh.gt.org>, John Hasler:
> [Snip...]
>>There is really no excuse for firmware of that sort hanging, though.
> I agree completely, and I've seen symptoms like this on only one other ISA
But most modems do. My three external modems all hang after a few
weeks. Modem banks hang. Yes, modem firmware is buggy. What's new?
> modem in several years, IIRC.
Then you haven't kept it up very long!
> BTW, I tried all kindsa "AT" reset combos I found while running it down (I
> don't recall exactly the commands, and apparently have no records now) but
> to no avail. On occasion, these suckers just seem to need a poweroff reset
> (AFAICT).
All modems do. It's called "my firmware is buggy and I just entered an
unforseen state from which there is no way out". Yes, nobody at the
factory has debiggedthe firmware, because it would take them several
weeks of watching and waiting too. And no, nobody does that.
Peter
THEOLDONE
U might also try here
http://home.att.net/~aubreyb/Linux.html
Go to page 2 and scroll down to modems
HTH
Robert Heller
"TB> #Harold Stevens US.972.952.3293 <ste...@mklog4.rsc.raytheon.com> wrote:
"TB> > In <87iswjo...@toncho.dhh.gt.org>, John Hasler:
"TB>
"TB> > [Snip...]
"TB>
"TB> >>There is really no excuse for firmware of that sort hanging, though.
"TB>
"TB> > I agree completely, and I've seen symptoms like this on only one other ISA
"TB>
"TB> But most modems do. My three external modems all hang after a few
"TB> weeks. Modem banks hang. Yes, modem firmware is buggy. What's new?
"TB>
"TB> > modem in several years, IIRC.
"TB>
"TB> Then you haven't kept it up very long!
"TB>
"TB> > BTW, I tried all kindsa "AT" reset combos I found while running it down (I
"TB> > don't recall exactly the commands, and apparently have no records now) but
"TB> > to no avail. On occasion, these suckers just seem to need a poweroff reset
"TB> > (AFAICT).
"TB>
"TB> All modems do. It's called "my firmware is buggy and I just entered an
"TB> unforseen state from which there is no way out". Yes, nobody at the
"TB> factory has debiggedthe firmware, because it would take them several
"TB> weeks of watching and waiting too. And no, nobody does that.
Here is a case of "they don't make them like they used to":
I have an *OLD* USR Robotics Courier 2400 modem. Made *before* 3Com
bought up USR. This modem still works. It has NEVER given me the least
bit of trouble. It is pushing 20 *years* old, in pretty much
*continious* service the whole time.
OTOH:
In less than 10 years I have gone though a USR Sportser 28.8K
(effectivly dies after less than a year or so), a USR Courier
V.Everything 33.6K (lasted maybe 7-8 years), and I am on my *second*
"Explorer V.90 Ex. (a couple of years for the first, < 6 months on the
current one).
"TB>
"TB> Peter
"TB>
Peter T. Breuer
> Here is a case of "they don't make them like they used to":
> I have an *OLD* USR Robotics Courier 2400 modem. Made *before* 3Com
Yeah, well, the courier is well known for excellence. The sportster is
the consumer model.
> bought up USR. This modem still works. It has NEVER given me the least
> bit of trouble. It is pushing 20 *years* old, in pretty much
> *continious* service the whole time.
> OTOH:
> In less than 10 years I have gone though a USR Sportser 28.8K
> (effectivly dies after less than a year or so), a USR Courier
> V.Everything 33.6K (lasted maybe 7-8 years), and I am on my *second*
> "Explorer V.90 Ex. (a couple of years for the first, < 6 months on the
> current one).
Probably.
Peter
Ray
Actiontec and USR make some good PCI hardware modems. Elsa (before they
died) made some USB hardware modems as does USR. Most machines have still
have serial ports so any external serial modem would work. There are
actually quite a few others, you just need to hunt around a bit.
--
Ray
Lee Sau Dan
Peter> All modems do. It's called "my firmware is buggy and I just
Peter> entered an unforseen state from which there is no way
Peter> out". Yes, nobody at the factory has debiggedthe firmware,
Peter> because it would take them several weeks of watching and
Peter> waiting too. And no, nobody does that.
Aren't there tools that would analyse the state transition graph to
detect such cycles? Or are the state transition functions really that
complicated for such a check? IOW, analytical check rather than
empirical check possible?
--
Lee Sau Dan 李守敦(Big5) ~{@nJX6X~}(HZ)
E-mail: dan...@informatik.uni-freiburg.de
Home page: http://www.informatik.uni-freiburg.de/~danlee
Lee Sau Dan
Robert> "Peter T. Breuer" <p...@oboe.it.uc3m.es>, In a message on
Robert> Mon, 20 Jan 2003 15:27:55 +0100, wrote :
TB> No, there are some (very few) pci hardware modems. You'll have
TB> to google for them - I forget the brand name.
Robert> And they tend to be more costly than an *external* RS232
Robert> modem.
Are there any technical reasons that they're more expensive? Are they
just more expensive because of rarity?
I used to like internal modems (ISA) because they were a bit cheaper
than external modems, and they don't require external power supply.
It's difficult for me to spare another mains socket for the included
transformer.
Fortunately, my last motherboard (a dual slot1 one indeed) still has a
few ISA slots. :) If not, I'd go for an external modem. Fewer
plug-and-pray hassles. Guanranteed to work under Linux (and all OSes,
because I still remember how to use AT commands and I can program).
Modems aren't that expensive nowadays anyway.
Peter T. Breuer
>>>>>> "Peter" == Peter T Breuer <p...@oboe.it.uc3m.es> writes:
> Peter> All modems do. It's called "my firmware is buggy and I just
> Peter> entered an unforseen state from which there is no way
> Peter> out". Yes, nobody at the factory has debiggedthe firmware,
> Peter> because it would take them several weeks of watching and
> Peter> waiting too. And no, nobody does that.
> Aren't there tools that would analyse the state transition graph to
If they had a finite state machine, yes. But most programmed designs
use stacks in some way (this is firmware), and trust they never run out
of memory for the stack. And anyway, nobody in a consumer electronics
firm has heard of formal methods, not even finite state analysis
techniques. And if they had heard, they wouldn't be writing the bios
code.
> detect such cycles? Or are the state transition functions really that
> complicated for such a check? IOW, analytical check rather than
> empirical check possible?
It's a good idea, and you have the proof that they don't do it. Or
can't do it. Whether it's actually impossible, I don't know, since
I don't know what's in their code. But I imagine they aren't using
a FSM generator.
Peter
John Hasler
> It's a good idea, and you have the proof that they don't do it. Or can't
> do it. Whether it's actually impossible, I don't know, since I don't know
> what's in their code. But I imagine they aren't using a FSM generator.
It's still possible to design the program so that the machine will always
recover, though it might reboot and lose state.
--
John Hasler
jo...@dhh.gt.org (John Hasler)
Dancing Horse Hill
Elmwood, WI
Peter T. Breuer
> Peter writes:
>> It's a good idea, and you have the proof that they don't do it. Or can't
>> do it. Whether it's actually impossible, I don't know, since I don't know
>> what's in their code. But I imagine they aren't using a FSM generator.
> It's still possible to design the program so that the machine will always
> recover, though it might reboot and lose state.
No .. that would also require a FSM analysis. They'd have to put a
reboot command in every unintended state, and by definition the
unintended states are the ones they don't know about, because they
didn't do any formal design or analysis ... basically we're talking
about 256KB of prom here, aren't we? That might in itself be pushing
it for control program analysis. How much ram do they have? That's
the real killer. The state's in there.
Peter
Robert Heller
In a message on 22 Jan 2003 08:52:04 +0100, wrote :
LSD> >>>>> "Robert" == Robert Heller <hel...@deepsoft.com> writes:
LSD>
LSD> Robert> "Peter T. Breuer" <p...@oboe.it.uc3m.es>, In a message on
LSD> Robert> Mon, 20 Jan 2003 15:27:55 +0100, wrote :
LSD>
LSD> TB> No, there are some (very few) pci hardware modems. You'll have
LSD> TB> to google for them - I forget the brand name.
LSD>
LSD> Robert> And they tend to be more costly than an *external* RS232
LSD> Robert> modem.
LSD>
LSD> Are there any technical reasons that they're more expensive? Are they
LSD> just more expensive because of rarity?
LSD>
Something like that, I believe. Rarity, plus the 'real' PCI modems
tend to be higher-end and have extra features. There is (unfortunatly)
no real (perceived) market for plain *real* PCI internal modems.
LSD> I used to like internal modems (ISA) because they were a bit cheaper
LSD> than external modems, and they don't require external power supply.
LSD> It's difficult for me to spare another mains socket for the included
LSD> transformer.
Not even with a cheap power strip?
LSD>
LSD> Fortunately, my last motherboard (a dual slot1 one indeed) still has a
LSD> few ISA slots. :) If not, I'd go for an external modem. Fewer
LSD> plug-and-pray hassles. Guanranteed to work under Linux (and all OSes,
LSD> because I still remember how to use AT commands and I can program).
LSD> Modems aren't that expensive nowadays anyway.
LSD>
LSD>
LSD> --
LSD> Lee Sau Dan 李守敦(Big5) ~{@nJX6X~}(HZ)
LSD>
LSD> E-mail: dan...@informatik.uni-freiburg.de
LSD> Home page: http://www.informatik.uni-freiburg.de/~danlee
LSD>
Pat D
Matthias Blankertz wrote:
> is there a list of which modems are winmodems and which models are real
> modems anywhere on the internet?
>
Finding a "real" modem can be a "real pain". A very partial list can be found on
http://www.vocpsystem.com/vgetty_modems.php as VOCP (a voice messaging system) uses vgetty which in
turn must use a working voice modem under Linux and other Unices.
You can perform a search for "vgetty voice modem" - if you find a brand that works with vgetty you
can be certain it a true modem with a controller. A lot (though certainly not all) of USR modems
work, including some internal 56k pci models (for instance the US Robotics 2976 56K V.90
Data/Fax/Voice has been reported to work)
HTH
--
Pat Deegan,
Registered Linux User #128131
http://www.psychogenic.com/contact.en.html
Lee Sau Dan
>> detect such cycles? Or are the state transition functions
>> really that complicated for such a check? IOW, analytical
>> check rather than empirical check possible?
Peter> It's a good idea, and you have the proof that they don't do
Peter> it. Or can't do it. Whether it's actually impossible, I
Peter> don't know, since I don't know what's in their code. But I
Peter> imagine they aren't using a FSM generator.
It's possible for FSM with no memory (FS automata). It's imposslbe
for a Turing machine with an infinitely long tape (i.e. unlimited
memory). For FSM with finite memory, the analysis should be possible.
Are you aware that the Java Virtual Machine is designed so that all
methods (i.e. functions in C) must manipulate the stack in a limited
way, so that you can apply algorithms to ensure that a given method is
well-behaved? E.g. every method is tagged with how many arguments
they take, and the method itself is not allowed to access (including
popping) the stack beyond its arguments and local variables. The
caller is restricted to pass the correct amount of arguments. It is
designed so that these checks can all be done in loading time, not in
run time. So, there is little runtime penalty (just once per load,
not once per execution) for such sanity checks. If all methods are
well-behaved, you can avoid problems like stack overflow or underflow.
I'm actually quite surprised that they don't use these methods for
those firmware. Or is user expectation really lower nowadays? I seem
to see more and more devices hanging and needing "reboots". Or just
our engineering skills are deteriorating?
Lee Sau Dan
John> It's still possible to design the program so that the
John> machine will always recover, though it might reboot and lose
John> state.
You mean some "watchdog" circuit, that would reset the other parts
when they appear to stop responding to external "stimuli"?
Peter T. Breuer
> John> It's still possible to design the program so that the
> John> machine will always recover, though it might reboot and lose
> John> state.
> You mean some "watchdog" circuit, that would reset the other parts
> when they appear to stop responding to external "stimuli"?
It doesn't work. They only have one cpu. But it's not a bad idea.
Peter
Peter T. Breuer
> >> detect such cycles? Or are the state transition functions
> >> really that complicated for such a check? IOW, analytical
> >> check rather than empirical check possible?
> Peter> It's a good idea, and you have the proof that they don't do
> Peter> it. Or can't do it. Whether it's actually impossible, I
> Peter> don't know, since I don't know what's in their code. But I
> Peter> imagine they aren't using a FSM generator.
> It's possible for FSM with no memory (FS automata). It's imposslbe
> for a Turing machine with an infinitely long tape (i.e. unlimited
> memory). For FSM with finite memory, the analysis should be possible.
Yes I KNOW. The problem is that if they write it as a normal program
they don't know how much memory it takes, and they'd have to prove an
upper bound before they even could start the analysis.
> Are you aware that the Java Virtual Machine is designed so that all
Thank you. Yes I AM. Apart from being about as closely connected to the
formal semantics of java as you can get, I am a theoretical computer
scientist whose speciality is programming language semantics and formal
methods.
> methods (i.e. functions in C) must manipulate the stack in a limited
> way, so that you can apply algorithms to ensure that a given method is
Sigh. It's not at all like that. Any stack machine is potentially
unbounded. Java can run out of memory. To do an FS analysis of any
program you first need to know how much memory it needs (to bound the
state space). If the JVM is running with other processes, then life is,
uh, complicated, since they can communicate by depriving each other of
memory, etc.
And then let's not get into the /size/ of the state space! Say 4MB
of memory. Uh, what IS 2 to the power of 32 million?
The best you can do is
a) either use model checking techniques after the fact, or
b) write a small tight FSM in the first place
> well-behaved? E.g. every method is tagged with how many arguments
> they take, and the method itself is not allowed to access (including
> popping) the stack beyond its arguments and local variables. The
You forget that it can call itself recursively. Try coding the ackerman
function.
> caller is restricted to pass the correct amount of arguments. It is
> designed so that these checks can all be done in loading time, not in
> run time. So, there is little runtime penalty (just once per load,
> not once per execution) for such sanity checks. If all methods are
> well-behaved, you can avoid problems like stack overflow or underflow.
No you can't. Obviously not. Just write
if (the riemann hypthesis is true)
then call myself until I overflow
else
don't.
and tell me what happens. Yes, the RH is checkable.
> I'm actually quite surprised that they don't use these methods for
Try telling them to do so. Their problem is that they are hiring
assembler programmers to write their bioses, not FSM experts. Or
anything like.
> to see more and more devices hanging and needing "reboots". Or just
> our engineering skills are deteriorating?
Most programmers don't know anything about what you just said.
Peter
James Silverton
"Robert Heller" <hel...@deepsoft.com> wrote in message
>>>>>>>>>>>>>>>>>>>Deletions<<<<<<<<<<<<
> LSD> I used to like internal modems (ISA) because they were a bit
cheaper
> LSD> than external modems, and they don't require external power
supply.
> LSD> It's difficult for me to spare another mains socket for the
included
> LSD> transformer.
>
> Not even with a cheap power strip?
>
>
It may be worth mentioning that there is one made by Curtis that has sockets
on three sides and allows using those monstrous voltage converters without
covering too many other sockets or tipping over.
--
James V. Silverton
Potomac, Maryland, USA
John Hasler
> It's possible for FSM with no memory (FS automata). It's imposslbe for a
> Turing machine with an infinitely long tape (i.e. unlimited memory). For
> FSM with finite memory, the analysis should be possible.
It's impractical for an FSM with a significant amount of memory, but
defensive design of both hardware and software can reduce the probability
of lockups to a negligible level.
> I'm actually quite surprised that they don't use these methods for those
> firmware.
The sort of firmware that goes into modems and similar devices is usually
written in a subset of C that can be cross-compiled for the DSPs and
microprocessors they use. Java would require more memory and a faster cpu
which would drive up the cost too much. It would also merely hide the
problems inside the JVM.
> I seem to see more and more devices hanging and needing "reboots". Or
> just our engineering skills are deteriorating?
The number of such devices being designed is expanding faster than the
supply of people capable of doing the work. The suits worsen the problem
by hiring certificates and buzzwords instead of competence.
John Hasler
> You mean some "watchdog" circuit...
Any non-trivial system in which a restart is not a critical failure should
have a watchdog. It doesn't make the system provably correct, of course,
but it eliminates lockups for practical purposes. There are other things
one can do as well.
> ...that would reset the other parts when they appear to stop responding
> to external "stimuli"?
Typically it resets the cpu if it fails to toggle one of its outputs often
enough.
> It doesn't work. They only have one cpu. But it's not a bad idea.
A watchdog doesn't have a cpu. It has an oscillator and some flipflops.
You can buy one for about $.15. Some cpus have them built in.
John Hasler
> If the JVM is running with other processes, then life is, uh,
> complicated, since they can communicate by depriving each other of
> memory, etc.
An embedded system that needs more than one process should use an RTOS
(yes, I know that does not solve the theoretical problem).
> And then let's not get into the /size/ of the state space! Say 4MB of
> memory. Uh, what IS 2 to the power of 32 million?
That's a lot of memory for a modem.
> Their problem is that they are hiring assembler programmers to write
> their bioses, not FSM experts.
C programmers, usually. Likely using an opaque DSP library supplied by the
DSP vendor and working for suits who figure it's ready to ship the day it
can dialup and connect successfully three times in a row.
Dave Brown
> Lee Sau Dan <dan...@informatik.uni-freiburg.de>,
> In a message on 22 Jan 2003 08:52:04 +0100, wrote :
>
> LSD> >>>>> "Robert" == Robert Heller <hel...@deepsoft.com> writes:
> LSD> Robert> And they tend to be more costly than an *external* RS232
> LSD> Robert> modem.
> LSD>
> LSD> Are there any technical reasons that they're more expensive? Are they
> LSD> just more expensive because of rarity?
>
>
> tend to be higher-end and have extra features. There is (unfortunatly)
> no real (perceived) market for plain *real* PCI internal modems.
I don't know about the "real market". Real PCI modems are priced about
the same as "real" ISA modems--about USD90. They have more than a couple
of signal-conditioning chips, including a real controller and DSP for
intelligent operation, therefore, more expensive.
I have a Zoom PCI faxmodem, (model 2920), which is a "real" PCI modem.
(Had to get it when I upgraded to a MB which had no ISA slots; I didn't
want to have a brick to power it, extra wires, etc.) BTW, bricks consume
power 24/7 even when the computer is off... not "environmentally
friendly". :)
--
Dave Brown Austin, TX
Robert Heller
In a message on Thu, 23 Jan 2003 20:53:19 GMT, wrote :
DB> In article <10432780...@cache1.news-service.com>, Robert Heller wrote:
DB> > Lee Sau Dan <dan...@informatik.uni-freiburg.de>,
DB> > In a message on 22 Jan 2003 08:52:04 +0100, wrote :
DB> >
DB> > LSD> >>>>> "Robert" == Robert Heller <hel...@deepsoft.com> writes:
DB> > LSD> Robert> And they tend to be more costly than an *external* RS232
DB> > LSD> Robert> modem.
DB> > LSD>
DB> > LSD> Are there any technical reasons that they're more expensive? Are they
DB> > LSD> just more expensive because of rarity?
DB> >
DB> >
DB> > Something like that, I believe. Rarity, plus the 'real' PCI modems
DB> > tend to be higher-end and have extra features. There is (unfortunatly)
DB> > no real (perceived) market for plain *real* PCI internal modems.
DB>
DB> I don't know about the "real market". Real PCI modems are priced about
DB> the same as "real" ISA modems--about USD90. They have more than a couple
DB> of signal-conditioning chips, including a real controller and DSP for
DB> intelligent operation, therefore, more expensive.
External RS232 Modems seem to go for about USD60, which is what my
Explorer V.90 Modem cost. This is a 56K, V.90, Voice/Fax modem.
I've seen 'WinModems' for under USD30.
DB>
DB> I have a Zoom PCI faxmodem, (model 2920), which is a "real" PCI modem.
DB> (Had to get it when I upgraded to a MB which had no ISA slots; I didn't
DB> want to have a brick to power it, extra wires, etc.) BTW, bricks consume
DB> power 24/7 even when the computer is off... not "environmentally
DB> friendly". :)
Unless you turn off the power strip...
DB>
DB> --
DB> Dave Brown Austin, TX
DB>
Lee Sau Dan
LSD> I used to like internal modems (ISA) because they were a bit
LSD> cheaper than external modems, and they don't require external
LSD> power supply. It's difficult for me to spare another mains
LSD> socket for the included transformer.
Robert> Not even with a cheap power strip?
Space is the main problem. It's so precious in Hong Kong.
--
Lee Sau Dan 李守敦(Big5) ~{@nJX6X~}(HZ)
E-mail: dan...@informatik.uni-freiburg.de
Home page: http://www.informatik.uni-freiburg.de/~danlee
Lee Sau Dan
>> It's possible for FSM with no memory (FS automata). It's
>> imposslbe for a Turing machine with an infinitely long tape
>> (i.e. unlimited memory). For FSM with finite memory, the
>> analysis should be possible.
Peter> Yes I KNOW. The problem is that if they write it as a
Peter> normal program they don't know how much memory it takes,
Peter> and they'd have to prove an upper bound before they even
Peter> could start the analysis.
Normal programs? That's simply not the right tool for the job.
>> well-behaved? E.g. every method is tagged with how many
>> arguments they take, and the method itself is not allowed to
>> access (including popping) the stack beyond its arguments and
>> local variables. The
Peter> You forget that it can call itself recursively. Try coding
Peter> the ackerman function.
Yeah. haha.. as an exercise, I did code the Ackerman function. But I
used dynamic programming instead.
But how would the modem firmware need real recursion? Are the
operations they perform really that complex?
And it isn't that hard to add some circuits (mainly a decoder) to
detect a stack overflow and trigger a reset.
>> If all methods are well-behaved, you can avoid
>> problems like stack overflow or underflow.
Peter> No you can't. Obviously not. Just write
Peter> if (the riemann hypthesis is true) then call myself until
Peter> I overflow else don't.
Is it impossible to do without recursion? Yes, you lose computation
power, but do you need that for a modulator-demodulator?
Peter> Try telling them to do so. Their problem is that they are
Peter> hiring assembler programmers to write their bioses, not FSM
Peter> experts. Or anything like.
Too bad. Like using VB to write a parser for C, instead of yacc.
Peter> Most programmers don't know anything about what you just
Peter> said.
Sigh. And I've seen programmers who can just do "HTML programming".
My first response was: "HTML programming" is not programming at all.
Peter T. Breuer
> >> It's possible for FSM with no memory (FS automata). It's
> >> imposslbe for a Turing machine with an infinitely long tape
> >> (i.e. unlimited memory). For FSM with finite memory, the
> >> analysis should be possible.
> Peter> Yes I KNOW. The problem is that if they write it as a
> Peter> normal program they don't know how much memory it takes,
> Peter> and they'd have to prove an upper bound before they even
> Peter> could start the analysis.
> Normal programs? That's simply not the right tool for the job.
But they don't know that. Even if they are implementing a finite state
machine, they will write it using normal programming techniques, and
then the program is not analysable, because the analysing tool has no
(generic) way of checking that it is intended to be a finite state
machine. They probably don't even bother to try. And if the
implementation is in any way "standard", it will /not/ be a FSM
(or at least not the one they intended - yes, everything is finite)
because there will be all kinds of interrupt conditions and out of
memory conditions that they never considered.
I.e. "bugs in the firmware".
> But how would the modem firmware need real recursion? Are the
It doesn't. But they write ina a language that /can do recursion/. It
doesn't matter if the program doesn't do it. That would have to be
proved. And recursion is no different from a while loop. Either would
require a theorm prover to analyse.
> And it isn't that hard to add some circuits (mainly a decoder) to
> detect a stack overflow and trigger a reset.
Sure, and that leaves you vulnerable to a condition that calls itself
in a tail recursion. No stack overflow. A simple "while(true);".
> Is it impossible to do without recursion? Yes, you lose computation
Don't fixate on recursion. It is the /power/ of the language that is
important. Unbounded loops, recursion, gotos, whatever. All give
the same mathematical power.
> power, but do you need that for a modulator-demodulator?
> Peter> Try telling them to do so. Their problem is that they are
> Peter> hiring assembler programmers to write their bioses, not FSM
> Peter> experts. Or anything like.
> Too bad. Like using VB to write a parser for C, instead of yacc.
Well, yes.
> Peter> Most programmers don't know anything about what you just
> Peter> said.
> Sigh. And I've seen programmers who can just do "HTML programming".
> My first response was: "HTML programming" is not programming at all.
Peter
Peter T. Breuer
> LSD> I used to like internal modems (ISA) because they were a bit
> LSD> cheaper than external modems, and they don't require external
> LSD> power supply. It's difficult for me to spare another mains
> LSD> socket for the included transformer.
> Robert> Not even with a cheap power strip?
> Space is the main problem. It's so precious in Hong Kong.
:-).
You can stick the power strip vertically up the wall, or horizontally
on the underside of the desk (which is what I do).
Peter
Lee Sau Dan
Dave> I have a Zoom PCI faxmodem, (model 2920), which is a "real"
Dave> PCI modem. (Had to get it when I upgraded to a MB which had
Dave> no ISA slots; I didn't want to have a brick to power it,
Dave> extra wires, etc.) BTW, bricks consume power 24/7 even when
Dave> the computer is off... not "environmentally friendly". :)
These are also reasons that I like internal modems better. I see that
brick redundant and a waste of space (which is so precious in my
environment) when I've already got a bigger brick in the computer
case. (In HK, I've seen some PCI/ISA "cards" or USB wirse which do
nothing but supply power to external devices. I've got a USB wire
that feeds power to my mobile phone.)
Too bad that we're getting more and more devices (not just modems)
that come with their own bricks. And they are only guaranteed to work
with the bundled bricks. Mobile phones, DV, digital cameras, a few
high energy consumption USB devices, ... It's nowadays not uncommon
to see a brick wall being built on a power bar base in a modern
office.
I think sooner or later, we will have just one big brick at home,
which would serve some standard voltage DC to all these brick-powered
equipments. We can then get rid of these bricks and plug the devices
directly to the wall sockets (of different shapes than AC sockets).
--
Lee Sau Dan 李守敦(Big5) ~{@nJX6X~}(HZ)
E-mail: dan...@informatik.uni-freiburg.de
Home page: http://www.informatik.uni-freiburg.de/~danlee
Lee Sau Dan
Peter> It doesn't. But they write ina a language that /can do
Peter> recursion/. It doesn't matter if the program doesn't do
Peter> it. That would have to be proved. And recursion is no
Peter> different from a while loop. Either would require a theorm
Peter> prover to analyse.
In theory, they have the same computation power. In practice, a loop
can be implemented in a more memory-efficient way than
non-tail-recursive recursion. For general recursion, you need a stack
to keep track of the return address. So, with limited memory, you
cannot have an arbitrarily large amounts of "iterations". With a
while loop, you can loop for arbitrarily many iterations and still
limit the memory needed to that needed by the loop variables and
loop-local variables.
Lee Sau Dan
John> The sort of firmware that goes into modems and similar
John> devices is usually written in a subset of C that can be
John> cross-compiled for the DSPs and microprocessors they use.
John> Java would require more memory and a faster cpu which would
John> drive up the cost too much. It would also merely hide the
John> problems inside the JVM.
I was not saying that the JVM be deployed there. I was just talking
about the technique of code verification. It's possible to restrict
to a set of "safe" instructions with which you can more easily check
(statically) whether you can end up with undesired results.
John> The number of such devices being designed is expanding
John> faster than the supply of people capable of doing the work.
John> The suits worsen the problem by hiring certificates and
John> buzzwords instead of competence.
Sigh.
Lee Sau Dan
>> Space is the main problem. It's so precious in Hong Kong.
Peter> :-).
Peter> You can stick the power strip vertically up the wall, or
Peter> horizontally on the underside of the desk (which is what I
Peter> do).
Every exploitable space is precious. Every gap is already stuffed
with (physical) files. Holes are used as pen holders, etc. Under the
desk? There's already a few boxes of other things: books, computer
accessories, etc., leaving only enough space for the feet. Up the
wall? There's already cupboards filled with stuff there! That's the
way to cope with a 300 sq ft flat (including washroom and kitchen!)
shared by 4 people. I'm now refraining from buying any more books,
because there would be no space to store them.
Peter T. Breuer
> >> Space is the main problem. It's so precious in Hong Kong.
> Peter> :-).
> Peter> You can stick the power strip vertically up the wall, or
> Peter> horizontally on the underside of the desk (which is what I
> Peter> do).
> Every exploitable space is precious. Every gap is already stuffed
> with (physical) files. Holes are used as pen holders, etc. Under the
> desk? There's already a few boxes of other things: books, computer
No, I said ON THE UNDERSIDE OF, not UNDER.
Peter
Peter T. Breuer
> Peter> It doesn't. But they write ina a language that /can do
> Peter> recursion/. It doesn't matter if the program doesn't do
> Peter> it. That would have to be proved. And recursion is no
> Peter> different from a while loop. Either would require a theorm
> Peter> prover to analyse.
> In theory, they have the same computation power. In practice, a loop
> can be implemented in a more memory-efficient way than
This is not important because you can't generally tell if the loop will use
memory or not (or if it will terminate ..). It becomes particularly
hard to tell if there are two threads of execution ...
> non-tail-recursive recursion. For general recursion, you need a stack
> to keep track of the return address. So, with limited memory, you
> cannot have an arbitrarily large amounts of "iterations". With a
> while loop, you can loop for arbitrarily many iterations and still
> limit the memory needed to that needed by the loop variables and
> loop-local variables.
The loop might do anything inside it, including pushing and popping a
stack. And if you don't allow pushing and popping (or malloc and free,
which amounts to the same thing), then you are in a FSM. But they
allowed it, because they used a general purpose language to write the
code .. assembler, in fact!
Anyway, the FSM is usually sufficiently large that you might as well be
in a general purpose programming language.
Peter
Kirk Strauser
Hash: SHA1
At 2003-01-22T07:52:04Z, Lee Sau Dan <dan...@informatik.uni-freiburg.de> writes:
> I used to like internal modems (ISA) because they were a bit cheaper than
> external modems, and they don't require external power supply. It's
> difficult for me to spare another mains socket for the included
> transformer.
I dislike internal modems, for the reason that they connect an external,
ungrounded power source (your phone line) directly to the bus on your
motherboard. When I worked tech support for a small ISP, we had 1-2 calls
after every thunderstorm that ended with "well, you have insurance, right?"
- --
Kirk Strauser
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.1 (GNU/Linux)
iD8DBQE+Lr7M5sRg+Y0CpvERAl3hAJ445vWQjTGq2rj0qU9j454j1F77PwCfXAdn
oZ1LgRZvZIJdvDk5G4n0snU=
=MTL+
-----END PGP SIGNATURE-----
John Hasler
> I dislike internal modems...
So do I.
> ...for the reason that they connect an external, ungrounded power source
> (your phone line)...
The phone line is grounded. The problem is that it is not grounded at the
same point nor with the same ground impedance as your power line.
> ...directly to the bus on your motherboard.
The phone line is isolated from the pc, though not always well enough.
Jean-David Beyer
>
> At 2003-01-22T07:52:04Z, Lee Sau Dan
> <dan...@informatik.uni-freiburg.de> writes:
>
>
>> I used to like internal modems (ISA) because they were a bit
>> cheaper than external modems, and they don't require external power
>> supply. It's difficult for me to spare another mains socket for
>> the included transformer.
>
>
> I dislike internal modems, for the reason that they connect an
> external, ungrounded power source (your phone line) directly to the
> bus on your motherboard. When I worked tech support for a small ISP,
> we had 1-2 calls after every thunderstorm that ended with "well, you
> have insurance, right?"
My telepone line is grounded somewhere, probably at the way into the
telephone company central office two miles from here. Possibly also at
the pole, though I doubt it. While there is a lightning protection
device in the little box outside my house, it probably does not ground
either pair in any way (two incoming lines).
I do have a surge protector on the telephone line that goes into the
computer, though. The surge protector is grounded to the UPS that powers
my computer, so they are within about one foot of one another. The whole
system's green wire goes to the water pipe that enters my house, and
also to a ground stake that goes about 8 feet into the ground (electric
code requires green wire to be grounded in two different places).
We get lightning storms here, but I have never lost a modem (yet).
--
.~. Jean-David Beyer Registered Linux User 85642.
/V\ Registered Machine 73926.
/( )\ Shrewsbury, New Jersey http://counter.li.org
^^-^^ 8:50am up 5 days, 15:01, 2 users, load average: 2.16, 2.28, 2.17
John Hasler
> While there is a lightning protection device in the little box outside my
> house, it probably does not ground either pair in any way (two incoming
> lines).
It does if it is properly installed.
> The surge protector is grounded to the UPS that powers my computer, so
> they are within about one foot of one another.
Good. That reduces the chance of a lightning spike driving the computer
and modem to different potentials, which is what does the damage.
> The whole system's green wire goes to the water pipe that enters my
> house...
More importantly, the neutral goes to the water line.
> ...and also to a ground stake that goes about 8 feet into the ground...
A ground rod can have as much as 25 ohms of resistance and typically has 8
or so. Might as well be an open circuit as far as lightning is concerned.
Of course, the risetimes of lightning spikes can be such that the
inductance of the wire from the service entrance to the water line can
develop several thousand volts across it even if your ground is a 280' well
casing as mine is.
> We get lightning storms here, but I have never lost a modem (yet).
I have, despite having an arrangement similar to yours. I also damaged a
motherboard.
--
John Hasler
jo...@dhh.gt.org (John Hasler)
Dancing Horse Hill
Elmwood, WI
Jean-David Beyer
> Jean-David Beyer writes:
>
>> While there is a lightning protection device in the little box
>> outside my house, it probably does not ground either pair in any
>> way (two incoming lines).
>
>
> It does if it is properly installed.
I really doubt that. The lightning protector in the old days was a
coupla carbon blocks that had fairly high resistance to the normal
signals in telephone lines, but relatively low resistance to the
voltages and currents expected from near-hits of lightning (they do not
try to protect against a direct hit, because that is hopeless).
Furthermore, in party-line days, you could signal three different
telephones because you put the ring between the two pair for one bell,
between one side and ground for the second, and between the other side
and ground for the third. They do not do that anymore, thank goodness,
but I doubt much else has changed, so no side of the phone line can be
grounded.
>
>
>> The surge protector is grounded to the UPS that powers my computer,
>> so they are within about one foot of one another.
>
>
> Good. That reduces the chance of a lightning spike driving the
> computer and modem to different potentials, which is what does the
> damage.
>
>
>> The whole system's green wire goes to the water pipe that enters my
>> house...
>
>
> More importantly, the neutral goes to the water line.
No it does not. It goes back to the transformer on the pole only. Of
course, it is grounded at the pole to a stake in the ground over 20 feet
from my house.
>
>
>> ...and also to a ground stake that goes about 8 feet into the
>> ground...
>
>
> A ground rod can have as much as 25 ohms of resistance and typically
> has 8 or so. Might as well be an open circuit as far as lightning is
> concerned.
They are only interested in safety. You should never have any current in
the green wire. Of course, you do, but it should be small unless there
is some kind of fault. Is the resistance in the water pipe surely less
than the 8-25 ohms you mention? I would not think it all that dependable
either.
> Of course, the risetimes of lightning spikes can be such
> that the inductance of the wire from the service entrance to the
> water line can develop several thousand volts across it even if your
> ground is a 280' well casing as mine is.
>
>
>> We get lightning storms here, but I have never lost a modem (yet).
>
>
> I have, despite having an arrangement similar to yours. I also
> damaged a motherboard.
--
.~. Jean-David Beyer Registered Linux User 85642.
/V\ Registered Machine 73926.
/( )\ Shrewsbury, New Jersey http://counter.li.org
^^-^^ 4:15pm up 5 days, 22:26, 3 users, load average: 3.23, 3.42, 3.51
Kirk Strauser
Hash: SHA1
At 2003-01-28T13:55:48Z, Jean-David Beyer <jdb...@exit109.com> writes:
> I do have a surge protector on the telephone line that goes into the
> computer, though. The surge protector is grounded to the UPS that powers
> my computer, so they are within about one foot of one another.
That's a great level of protection for a normal power surge, but probably
won't do much to stop a lightning strike. Imagine a nice, fat fuse
protection your line. Now, imagine that the potential across that fuse goes
- From 120V to 1e+6V within a couple of microseconds. First, the fuse may or
may not blow in time. Second, if it *does* blow in time, the resulting
metal gas may be conductive enough to let through more voltage that you'd
want to test with your tongue.
> The whole system's green wire goes to the water pipe that enters my house,
> and also to a ground stake that goes about 8 feet into the ground
> (electric code requires green wire to be grounded in two different
> places).
See above. How long do you think it'd take for that ground wire to melt
under a direct lightning strike?
> We get lightning storms here, but I have never lost a modem (yet).
...thereby guaranteeing that it'll happen within the next month. ;-)
Seriously, surge protectors are great, but the ones you're likely to buy for
residential use may or may not save your equipment. Are they better than
nothing? Absolutely. Are they foolproof? Not by a longshot.
- --
Kirk Strauser
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.0 (GNU/Linux)
iD8DBQE+Nun65sRg+Y0CpvERAvAPAJ9r6vdBkDLQ6ss7IuOa9CTAeBim4ACcDgCS
gNX1RH15iM6QoyGU8x945Cw=
=Hd7z
-----END PGP SIGNATURE-----
John Hasler
> The lightning protector in the old days was a coupla carbon blocks that
> had fairly high resistance to the normal signals in telephone lines, but
> relatively low resistance to the voltages and currents expected from
> near-hits of lightning (they do not try to protect against a direct hit,
> because that is hopeless).
You're right: neither side is directly grounded at the house.
> No it does not. It goes back to the transformer on the pole only.
If the neutral is not grounded at the service entrance your house is
miswired and not in compliance with the NEC.
> Is the resistance in the water pipe surely less than the 8-25 ohms you
> mention?
Depends. If the pipe has no dielectric unions and is electrically
continuous all the way to the main it's ground resistance is likely to be
very low.
> I would not think it all that dependable either.
Yes. Plumbers are not electricians and are under no obligation to maintain
electrical continuity. Your best available ground is probably the
reinforcing rod in your basement floor (if any).
John Hasler
> How long do you think it'd take for that ground wire to melt under a
> direct lightning strike?
If lightning hits the pole-pig next to your house the best you can hope for
is that the place won't burn down. However, if it hits the transmission
line a mile down the road proper bonding and grounding might be enough to
protect your equipment.
w_tom
posted here. Not only must the surge protector connect to
central earth ground BUT it must make a 'less than 10 foot'
connection. 10 foot is defined by fundamental electrical
concepts. That outlet safety ground does eventually connect
to earth ground for low frequency AC electric - to provide
human safety. But that same wire is too long for surges.
For human safety, wire 'resistance' is critical. For
transistor safety, wire 'impedance' is critical. A plug-in
surge protector has all but no earth ground - too much
impedance. Effective surge protection must have a dedicated,
'less than 10 foot' connection to central earth ground.
This technology is so old and so well proven that residental
phone lines have effective 'whole house' surge protection
within 10 feet of central earth ground. That telco line
'whole house' surge protector is installed free for protection
even from direct lightning strikes. However the naive
recommend plug-in surge protectors because they 1) don't know
of existing surge protector, 2) don't learn about the critical
less than 10 foot connection to central earth ground, and 3)
don't even understand how modems are damaged.
A surge protector is only as effective as its earth ground.
Phone lines have effective surge protection installed by the
telco. But plug-in surge protectors have all but no earth
ground. No earth ground means no effective surge protection.
So instead, those plug-in manufacturers just forget to mention
anything about earthing to sell ineffective products at
inflated prices.
There is no significant difference between external or
internal modems. Both are connected directly to AC safety
ground. However that safety ground can become the incoming
source of surges. Why? The technical facts are discussed in
great detail in discussions in the newgroup misc.rural
Storm and Lightning damage in the country 28 Jul 2002
Lightning Nightmares!! 10 Aug 2002
A fundamental fact. A surge protector is only as effective
as its earth ground. That means a plug-in surge protector
cannot be effective having all but no earth ground.
Jean-David Beyer wrote:
w_tom
That ground wire is so effective that direct strike damage is
routinely avoided since before WWII. FM and TV electronics
atop the Empire State Building are struck directly 25 times
per year. Damage is not even an option. If direct strikes
were not earthed without damage by ground wires, then
telephone and 911 emergency operators would periodically be
taken out dead from direct strikes to headsets. Telco
switching centers connected to overhead wires everywhere in
town would shut down with every T-storm to avoid
$multi-million damage. These things don't happen because a
ground wire is sufficient to earth direct strikes.
Common knowledge is that well over 90% of direct lightning
strikes even to trees don't even leave appreciable marks.
Properly sized earth grounding wires also are not damaged by
direct strike. But that wire must meet certain basic
electrical requirements. Number 1 requirement - it must be
short to earth ground - the necessary 'less than 10 foot'
connection.
In the meantime, that UPS does not even claim to provide
essential common mode surge protection. Ineffective surge
protectors are defined by any of four characteristics: 1) too
close to transistors, 2) too far from earth ground, 3) too
expensive to only protect one appliance, and 4) too undersized
(too few joules). Many plug-in UPSes meet all four
characteristics of ineffective protection.
Jean-David Beyer
> Kirk Strauser writes:
>
>> How long do you think it'd take for that ground wire to melt under
>> a direct lightning strike?
>
>
> If lightning hits the pole-pig next to your house the best you can
> hope for is that the place won't burn down. However, if it hits the
> transmission line a mile down the road proper bonding and grounding
> might be enough to protect your equipment.
Right. For a direct hit, about the only things that will save you are
not being there at the time, making sure your insurance is paid up, and
that your full backup tapes are safely in your safety deposit box at a
bank some distance away.
--
.~. Jean-David Beyer Registered Linux User 85642.
/V\ Registered Machine 73926.
/( )\ Shrewsbury, New Jersey http://counter.li.org
^^-^^ 10:50pm up 6 days, 5:01, 2 users, load average: 2.47, 2.49, 2.27
Lee Sau Dan
Peter> The loop might do anything inside it, including pushing and
Peter> popping a stack. And if you don't allow pushing and
Peter> popping (or malloc and free, which amounts to the same
Peter> thing), then you are in a FSM. But they allowed it,
Peter> because they used a general purpose language to write the
Peter> code .. assembler, in fact!
That's the problem.
BTW, back in undergrad final year, a classmate did a project with a
DSP card. It came with some development tools and specs. The tools
contain some demo programs, which are structed well. Although they're
in C, it's not difficult upon reading the code to discover that it is
styled as a finite state machine. To extend or modify those examples
is quite easy once you realize that it's just a set of states. You
then copy a template state, modify it to your own needs, and then
compile!
Peter> Anyway, the FSM is usually sufficiently large that you
Peter> might as well be in a general purpose programming language.
General enough to hang easily, let alone the possibility of a virus
infection. :)
Lee Sau Dan
Peter> You can stick the power strip vertically up the wall, or
Peter> horizontally on the underside of the desk (which is what I
Peter> do).
>> Every exploitable space is precious. Every gap is already
>> stuffed with (physical) files. Holes are used as pen holders,
>> etc. Under the desk? There's already a few boxes of other
>> things: books, computer
Peter> No, I said ON THE UNDERSIDE OF, not UNDER.
There is no more space there. Things stuffed under the desk are
already touching the underside of the desk.
Kirk Strauser
Hash: SHA1
At 2003-01-29T03:32:52Z, w_tom <w_t...@hotmail.com> writes:
> This poster demonstrates a lack of real world experience.
By this, you mean "you"?
Look, I'm not making this stuff up. As I said earlier, I've personally
talked to 15-20 people whose computers were burnt to a crisp via lightning
into an internal modem. I'm certain that not every single one of them lived
in old houses with ancient wiring. A good friend, for instance, lived in a
10-year-old home, used surge protectors, and still lost hardware.
I'm sure that you know something about the subject, but I've seen this
happen first-hand, and I'm not particularly interested in someone telling me
that it's either impossible or extremely unlikely.
To re-iterate, when I worked in tech support, it was well-accepted that a
thunderstorm *would* result in one or two callers who were online during the
storm, and now can't boot their computers. Guaranteed, every time.
- --
Kirk Strauser
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.1 (GNU/Linux)
iD8DBQE+OARW5sRg+Y0CpvERAhW4AJ4lImYLfSbBp7Pho+kSh3qasj7OqgCcCmrr
80rfcwfaJio9h7QYorJ20pU=
=2y3p
-----END PGP SIGNATURE-----
w_tom
understand both theory and experimental evidence to have
knowledge. Have you built surge protectors, seen them work or
fail, and learned the underlying theory? I have even
repaired modems by replacing components on the PC board - to
learn how surges damage modems. I learned the circuit paths
used by surges through modems by replacing the defective
parts.
Your friend used surge protectors and still lost modems.
Of course. Was the surge protector at the service entrance
and connected short to central earth ground? If not, then the
surge protector many even have contributed to his modem
damage. Just because it is called a surge protector does not
mean it provides surge protection. That assumption is called
junk science.
It is called 'whole house' surge protection. Principles
were proven long before either you or I existed. A surge
protector is only as effective as its earth ground - as even
Ben Franklin demonstrated in 1752. To appreciate why surge
protectors work successfully, one must understand the
difference between impedance and resistance, AND must
understand why that concept is appropriate to surge
protection. These are 1st year EE concepts.
Cited previously is a discussion in newsgroup misc.rural.
Nothing technically advanced. But if simple concepts are not
understood, then the reader does not have sufficient knowledge
to know even what a surge protector does. Demonstrated is why
plug-in surge protectors are not effective (safety ground wire
does or doesn't exist) and why a surge protector must connect
less than 10 foot to central earth ground.
A tech support person should be familiar with why surge
damage happens. Figure from an industry professional shows how
effective surge protection is installed - the essential single
point ground for each structure. Destructive surges can even
enter on underground wires. Notice no plug-in surge
protectors:
http://www.erico.com/erico_public/pdf/fep/TechNotes/Tncr002.pdf
A benchmark in surge protection is Polyphaser. Tech support
should know that name. Polyphaser's application notes are
considered legendary by industry professionals. What do they
discuss? Earthing, because surge protection is about earthing
a surge - not about surge protectors:
http://www.polyphaser.com/ppc_technical.asp
http://www.telebyteusa.com/primer/ch6.htm
> Section 6.4:
> Conceptually, lightning protection devices are switches to
> ground. Once a threatening surge is detected, a lightning
> protection device grounds the incoming signal connection point
> of the equipment being protected. Thus, redirecting the
> threatening surge on a path-of-least resistance (impedance)
> to ground where it is absorbed.
http://www.ipclp.com/html/aud_ho_faq.html
> A properly installed lightning protection system intercepts
> the lightning bolt between cloud and earth and harmlessly
> conducts it to ground without damage.
> Yes, in addition to the lightning protection system consisting
> of air terminals, conductor cables, clamps, fasteners, 10 foot
> grounds, etc., a secondary lightning suppressor is installed
> on your electric service entrance panel to prevent current
> fluctuations (called lightning surges) during a thunderstorm.
That secondary lightning suppressor is the 'whole house'
surge protector.
Does the surge enter on phone line, destroy a modem, then
stop? Of course not. Surges are electricity. No surge
damage occurs if a complete circuit does not exist. An
incoming and outgoing path must exist through that modem to
have surge damage. One does not have a clue how a surge
damaged something until one first defines both that incoming
(from cloud) and outgoing (to earth) circuit path. This was
defined in the above cited discussions you were expected to
read in misc.rural.
Industry professionals learn underlying principles which is
why both internal and external modems are exposed to same
surge damage.
Again, you don't have the real world experience. You have
heard about 15 symptoms. But did you follow the complete
circuit - incoming and outgoing through the building? Did you
replace damaged components to make the modem work? Does you
experience also demonstrate that destructive surges occur
typically once every eight years?
Did you first learn from IEEE papers that demonstrate that
outlet safety ground wire, if used for surge protection, would
induce surges on all other household wires? Read that
sentence again. That ground wire bundled with other household
wires would induce surges throughout the building if a surge
was using that safety ground wire. Again, another reason why
a surge protector to earth ground wire must be less than 10
feet - and other essential fundamentals defined in those
misc.rural newsgroup discussions.
Those posts in misc.rural were not read which explains so
little knowledge about surge damage. Those who have both
learned theory and have extensive experience know why 'whole
house' surge protectors with a short connection to earth
ground are critical. Don't response yet. First learn how
much there is to know about surge protection AND how
ineffective plug-in UPSes and surge protectors really are. You
have many hours of reading in those two discussion in
misc.rural. Many industry sources cited. The importance of
this sentence will then be important: a surge protector is
only as effective as its earth ground.
w_tom
strike will melt ground wires. That conclusion is classic
junk science. Numbers are not even provided.
From Colin Baliss "Transmission & Distribution Electrical
Engineering":
> Although lightning strikes have impressive voltage and current
> values (typically hundreds to thousands of kV and 10-100 kA)
> the energy content of the discharge is relatively low and most
> of the damage to power plant is caused by 'power follow-through
> current'.
Read again. The energy content of the discharge is
relatively low which is why wires are not vaporized by typical
lightning strikes. Dr Martin A Uman makes the same point:
> Most of the energy available to the lightning is converted
> along the lightning channel to thunder, heat, light, and
> radio waves, leaving only a fraction available at the channel
> base for immediate use or storage.
Lightning is not the high energy event of urban myth. But
then urban myth purveyors make conclusions without numbers.
Myths are from wild speculation. Surge damage is routinely
avoided by grounding wires.
If a direct strike vaporized wires, then even Ben Franklin's
lightning rods would have failed. Jean-David Beyer has posted
a classic example of junk science reasoning. He has posted in
direct contradiction even to 1930s Westinghouse and GE
papers. Surge damage from direct strikes is so easily
avoidable as to be directly traceable to human failure. Only
myth purveyors will claim massive energy at the lightning
strike point. Surge damage from a direct strike is directly
traceable to human failure.
Kirk Strauser
Hash: SHA1
At 2003-01-29T19:51:54Z, w_tom <w_t...@hotmail.com> writes:
> As demonstrated by high school science, one must understand both theory
> and experimental evidence to have knowledge. Have you built surge
> protectors, seen them work or fail, and learned the underlying theory?
Nope. I've just seen what happens when they fail. I'm not really
interested in arguing about it, so I leave you to think as you may. Have a
nice day.
- --
Kirk Strauser
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.1 (GNU/Linux)
iD8DBQE+OEBL5sRg+Y0CpvERAlYMAJ9netzV+NOP3e41xXyAKmoa/7/LtwCeIeCl
nZDilIWiNu2uUgowBXVd9gc=
=nZ04
-----END PGP SIGNATURE-----
w_tom
during a surge. A surge protector that fails leaves
electronics exposed to the rest of a surge and to all future
surges. Failed surge protectors is how a naive public
recommends such ineffective devices to friends - undersized
devices sold at excessive profit.
Characteristic of effective 'whole house' surge protectors.
They shunt a direct lightning strike to earth ground AND do
not fail. Effective surge protectors start at about 50,000
amps and remain functional for the next surge. Ineffective
(undersized) surge protectors even fail on surges not
sufficient to damage adjacent electronics. Then the naive
recommend such undersized protectors to friends - and buy
more. Go figure. What a racket - to get the naive to buy
even more ineffective, undersized, and grossly overpriced
surge protectors.
A failed surge protector is one that provided ineffective
surge protection. Grossly undersized surge protectors even
fail on surges insufficient to damage computers. But
unfortunately, those without sufficient background knowledge
call these grossly undersized devices "effective protection"
because they assume.
Surge protectors that fail did not provide effective surge
protection. Then there is the missing earth ground. A surge
protector is only as effective as its earth ground.
bill davidsen
Peter T. Breuer <p...@oboe.it.uc3m.es> wrote:
| Lee Sau Dan <dan...@informatik.uni-freiburg.de> wrote:
| > John> machine will always recover, though it might reboot and lose
| > John> state.
|
| > You mean some "watchdog" circuit, that would reset the other parts
| > when they appear to stop responding to external "stimuli"?
|
| It doesn't work. They only have one cpu. But it's not a bad idea.
What has the number of CPUs to do with anything? Watchdog circuits
reboot the machine. Software watchdogs *usually* reboot the machine.
Neither depend on the number of CPUs.
--
Bill Davidsen <davi...@tmr.com> CTO, TMR Associates
A hero is not someone who only does the brave thing when challenge is
thrust upon him. A hero chooses to accept personal risk for the common
good when it could be avoided; to slay the tiger, extinguish the
burning building, or venture into space.