networking with xp
Carlo
I loaded Openserver 5.04 on a machine at home. It has a Realtek
network card in it which is connected to another Realtek card in a
machine running XP Pro (crossover cable)
The IP address for the Unix machine is 199.1.1.10 (set up in
scoadmin).
In etc/hosts there are two entries: 127.0.0.1 and 199.1.1.10. (How do
I clear a line totally in VI?)
If I ping 199.1.1.10 it does not lose any packets. Is this a sure sign
that the network card is configured correctly - I set it up as IBM
NE2000 compatible - the i/o address is 300 and i/vector is 3 with no *
to indicate a conflict.
I then set up the XP machine with an IP address of 199.1.1.1 (prior to
loading Openserver the machine was loaded with 98SE and the network
was up and fine - different IP address). If I ping 199.1.1.1 on the XP
machine it fails:
Network Adapters [00000009] Realtek RTL8139 Family PCI Fast Ethernet
NIC FAILED
Caption = [00000009] Realtek RTL8139 Family PCI Fast Ethernet NIC
DatabasePath = %SystemRoot%\System32\drivers\etc
Description = Realtek RTL8139 Family PCI Fast Ethernet NIC
DHCPEnabled = FALSE
DNSEnabledForWINSResolution = FALSE
DNSHostName = CARLO_XP
Index = 9
IPAddress = 199.1.1.1 (FAILED)
Pinging 199.1.1.1 with 32 bytes of data:
Request timed out.
Request timed out.
Request timed out.
Request timed out.
Ping statistics for 199.1.1.1:
Packets: Sent = 4, Received = 0, Lost = 4 (100% loss)
IPConnectionMetric = 20
IPEnabled = TRUE
IPFilterSecurityEnabled = FALSE
IPSecPermitIPProtocols = 0
IPSecPermitTCPPorts = 0
IPSecPermitUDPPorts = 0
IPSubnet = 255.255.255.0
IPXEnabled = FALSE
MACAddress = 00:50:AB:00:07:34
ServiceName = RTL8139
SettingID = {88F1AA4B-2319-40F9-A97B-9551DD22F5B2}
TcpipNetbiosOptions = 0
WINSEnableLMHostsLookup = TRUE
and when I try ping 199.1.1.1 from the unix machine I get 'host is
down'.
Firstly, is the problem only on the XP side?
Is there anything else I need to set up on the Unix side to get the
network to talk. (I have an application on the Openserver side that I
need to basically telnet into from the XP side to run it)
I have been trying for some days now with no success. Please help.
Thanks
Carlob
Tony Lawrence
> Hi
> I loaded Openserver 5.04 on a machine at home. It has a Realtek
> network card in it which is connected to another Realtek card in a
> machine running XP Pro (crossover cable)
Why use cheap crap like Realtek in a server?
> The IP address for the Unix machine is 199.1.1.10 (set up in
> scoadmin).
Bad choice. See http://aplawrence.com/Unixart/net101.html
> In etc/hosts there are two entries: 127.0.0.1 and 199.1.1.10. (How do
> I clear a line totally in VI?)
See http://aplawrence.com/Unixart/viprimer.html
But the 127 line belongs there - don't remove it.
> If I ping 199.1.1.10 it does not lose any packets. Is this a sure sign
> that the network card is configured correctly
No. It's meaningless. That ping never even gets to the hardware.
Trying to test this with only two machines is very difficult. Your
problem could be in either machine or your cable.
(stuff deleted)
> and when I try ping 199.1.1.1 from the unix machine I get 'host is
> down'.
> Firstly, is the problem only on the XP side?
Maybe - who knows?
> Is there anything else I need to set up on the Unix side to get the
> network to talk. (I have an application on the Openserver side that I
> need to basically telnet into from the XP side to run it)
> I have been trying for some days now with no success. Please help.
You need a third device here and a hub or a switch to keep your sanity.
Doesn't need to be a machine: a print server or a cable/dsl router
would be fine - you just need something more than two machines to chjeck
against.
--
Tony Lawrence
Free SCO and Linux Skills Tests: http://aplawrence.com/skillstest.html
Dan Skinner
Carlob;
I guess you learned the lesson we all have to learn once. It's better
to break only one thing at a time.
I take it that you can ping the SCO machine from its own console.
That's a good
thing but not enough. You should be able to telnet to yourself from
the SCO console (telnet 199.1.1.10) That would also be a good sign.
It says a lot of network stuff is working, but does not say the nic is
working fully.
If I were you I would get the windows machine to the same place. That
is it can ping itself without errors.
Then if you have a friend with a hub or switch and a couple of
straight thru cat5 jumpers I'd borrow them and put them in your
network. The connection lights on the hub/switch will tell a lot
about the status of the nic's and the network. If you have good
lights then try to ping one machine form the other.
Look for results on the screen and for traffic on the lights. If that
same friend has a working machine on that network you might want to
borrow that also
and check out the lights and ping your machines from it.
Once your network works with the hub/switch try the crossover cable.
It should work if it is made right, but it doesn't give any signs
except smoke (and then you're in real trouble!!).
Good luck
Regards...Dan.
Tom Parsons
Step #1: Open case on each computer
Remove the Realtek NIC
Apply 16 lb sledge hammer to each card
Deposit remains in appropriate receptable
Buy a pair of reliable, quality network cards. You've already
spent the difference in price fiddling with this junk.
Install the better cards and go to Step #2
Step #2: Are you really "names.wjp.net"? If not, don't hijack their
ip address(es).
These blocks of IP addresses are reserved for private, non-internet
connnected networks. Use them.
10.0.0.0 to 10.255.255.255
172.16.0.0 to 172.16.255.255
192.168.0.0 to 192.168.255.255
With quality cards and proper ip addresses, your problems may just disappear.
--
==========================================================================
Tom Parsons t...@tegan.com
==========================================================================
Dan Skinner
> Carlo enscribed:
> | Hi
<snip>
Carlo;
Tony and Tom seem more interested in commenting on the quality of
digital
equipment, than on trying to help you out of the perdickerment.
I agree that you have not selected top line stuff, but i assume your
planned use is limited. You have selected in-appropriate ip's but
since you connected them with a crossover cable the chance of internet
damage is limited.
If they are new and you can return them, I would. I find that it
usually pays to use SCO certified parts in SCO box. They are not
really better, but they are known to work. The windows end doesn't
matter as much, but while you are buying one, buy two. While I'm
spending your money, buy a little, cheap switch, built into a
cable/dsl router if you are looking forward to broadband. (Linksys so
Tony and Tom can flame me) I have an SCO 3.2v4.2 desktop, an SCO
3.2v5.0.4 server, and a Linux 2.4.8-26mdk
Dan Skinner
OOPS, didn't brag about my typing skills did I?
Carlo; Tony and Tom seem more interested in commenting on the quality
of digital equipment, than on trying to help you out of the
perdickerment. I agree that you have not selected top line stuff, but
I assume your planned use is limited. You have selected in-appropriate
ip's but since you connected them with a crossover cable the chance of
internet damage is limited. If they are new and you can return them, I
would. I find that it usually pays to use SCO certified parts in SCO
box. They are not really better, but they are known to work. The
windows end doesn't matter as much, but while you are buying one, buy
two. While I'm spending your money, buy a little, cheap switch, built
into a cable/dsl router if you are looking forward to broadband.
(Linksys so Tony and Tom can flame me) I have an SCO 3.2v4.2 desktop,
an SCO 3.2v5.0.4 server, and a Linux 2.4.8-26mdk server plus a half
dozen Microsoft machines connected by Linksys switch and cable router
to internet by cable modem (Charter.) (We do software development and
support.) Some of the nics are old (1993) some are new, some are
wireless, some I paid $4.00 for ($4.00 would be new one). In my
experience high priced gear has a slightly lower DOA (dead on arrival)
rate and may last a little longer. While they are working they all
work about the same.
Buy the parts and call it trouble shooting expense. Frame the
crossover cable with a plaque that reads "It shoulda worked".
Anyway, good luck and don't give up!
Regards…Dan.
Tony Lawrence
> Tom Parsons <c...@tegan.com> wrote in message news:<2002122617...@tegan.com>...
>
>>Carlo enscribed:
>>| Hi
>
> <snip>
> Carlo;
> Tony and Tom seem more interested in commenting on the quality of
> digital
> equipment, than on trying to help you out of the perdickerment.
Excuse me?
BTW, it's predicament, but the advice I gave him after commenting on the
card was no different than what you gave him.
> I agree that you have not selected top line stuff, but i assume your
> planned use is limited. You have selected in-appropriate ip's but
> since you connected them with a crossover cable the chance of internet
> damage is limited.
That's not the point. The reason Tom and I mentioned that is that this
person will go on to other places and continue to use inappropriate ip
addresses. Also, just the posting of those addresses can cause other
people to copy them for their own networks, ultimately causing more
trouble for people like us who clean up messes like that regularly.
> If they are new and you can return them, I would. I find that it
> usually pays to use SCO certified parts in SCO box. They are not
> really better, but they are known to work.
Oh, foo. You don't have a clue what you are talking about. It's not a
matter of "SCO Certified", it's a matter of being low price, low
performance, poorly engineered junk. These low end cards cause grief
everywhere. Sometimes they'll work ok on a local lan but the minute
they have to go through a router they stop working. These things are
regularly the cause of intermittent, flaky and "unexplainable" network
behavior.
> The windows end doesn't
> matter as much,
Yes, it does, and for the same reasons.
>but while you are buying one, buy two. While I'm
> spending your money, buy a little, cheap switch, built into a
> cable/dsl router if you are looking forward to broadband. (Linksys so
> Tony and Tom can flame me) I have an SCO 3.2v4.2 desktop, an SCO
> 3.2v5.0.4 server, and a Linux 2.4.8-26mdk
There's nothing wrong with Linksys for a home setup. I use a similar
device of a different brand here in my own home. I even have small
clients with Linksys routers. I wouldn't use it in a larger setup, but
there's nothing wrong with it if properly configured (like not leaving
the default password which I see in so many places).
Dan Skinner
> Dan Skinner wrote:
> > Tom Parsons <c...@tegan.com> wrote in message news:<2002122617...@tegan.com>...
> >
> >>Carlo enscribed:
> >>| Hi
> >
> > <snip>
>
> BTW, it's predicament, but the advice I gave him after commenting on the
> card was no different than what you gave him.
Which goes to prove that you are not without redeeming value.
>
> > I agree that you have not selected top line stuff, but i assume your
> > planned use is limited. You have selected in-appropriate ip's but
> > since you connected them with a crossover cable the chance of internet
> > damage is limited.
>
> That's not the point. The reason Tom and I mentioned that is that this
> person will go on to other places and continue to use inappropriate ip
> addresses. Also, just the posting of those addresses can cause other
> people to copy them for their own networks, ultimately causing more
> trouble for people like us who clean up messes like that regularly.
Then the goal was to educate and inform. The condescending attitude
had me fooled.
I assume you do it for a living. Without the errors of beginners, you
would have less work
(Read: less income.) I learned many things at IBM instructor training
in the 60's . Two of them are: 1:( You'll motivate more with honey
than with vinegar) 2:( If the student knows more than you do, call him
sir and listen.)
>
> > If they are new and you can return them, I would. I find that it
> > usually pays to use SCO certified parts in SCO box. They are not
> > really better, but they are known to work.
>
> Oh, foo. You don't have a clue what you are talking about. It's not a
> matter of "SCO Certified", it's a matter of being low price, low
> performance, poorly engineered junk. These low end cards cause grief
> everywhere. Sometimes they'll work ok on a local lan but the minute
> they have to go through a router they stop working. These things are
> regularly the cause of intermittent, flaky and "unexplainable" network
> behavior.
That explains what happened to me a few months ago. I installed a
couple of hi-end
3com (SCO certified) cards in a windows machine cause they were in
inventory. They would not work! I sent my assistant to Best Buy for
a couple of the $4.00 cards ( OK $4.99.) They worked right out of the
box. The 3com cards were later installed in an SCO box and worked
fine. For $10.00 I can't afford to find out why, but the facts are
the facts. The Windows machines were assembled and tested using
Linksys router access to the internet. They were later joined by the
SCO box behind a Cisco router of some type at the customer site. I
think this proves your point.
The 3com cards are big, heavy and have lots of surface components.
The $4.00 cards are small, light and have very few surface components.
Again your point is proved. The 3com cards were priced at $69.00 and
the $4.00 cards were priced at $4.99. The 3com cards run at 10mhz the
$4.00 cards at 100mhz. Ain't engineering wonderful?
Oh and Tony, do you have a theory why the bits get confused when they
pass through a router?
>
>
> > The windows end doesn't
> > matter as much,
>
> Yes, it does, and for the same reasons.
>
>
>
> >but while you are buying one, buy two. While I'm
> > spending your money, buy a little, cheap switch, built into a
> > cable/dsl router if you are looking forward to broadband. (Linksys so
> > Tony and Tom can flame me) I have an SCO 3.2v4.2 desktop, an SCO
> > 3.2v5.0.4 server, and a Linux 2.4.8-26mdk
>
> There's nothing wrong with Linksys for a home setup. I use a similar
> device of a different brand here in my own home. I even have small
> clients with Linksys routers. I wouldn't use it in a larger setup, but
> there's nothing wrong with it if properly configured (like not leaving
> the default password which I see in so many places).
And Tony, you leave your Cisco routers installed with default
settings?
I'll take the Linksys stuff home as soon as the bits show signs of
confusion.
Keep up the good work. We enjoy your good advice, even if it's framed
in bad politics.
Regards…Dan.
Rainer Zocholl
>Carlo enscribed:
>| Hi
>| I loaded Openserver 5.04 on a machine at home. It has a Realtek
>| network card in it which is connected to another Realtek card in a
>| machine running XP Pro (crossover cable)
>| The IP address for the Unix machine is 199.1.1.10 (set up in
>| scoadmin).
>Step #1: Open case on each computer
> Remove the Realtek NIC
> Apply 16 lb sledge hammer to each card
> Deposit remains in appropriate receptable
An oecoligical alternative: Spent it your "best" enemy...
(Before the sledge hammer optimization)
>Step #2: Are you really "names.wjp.net"? If not, don't hijack their
> ip address(es).
> These blocks of IP addresses are reserved for private,
>non-internet connnected networks. Use them.
> 10.0.0.0 to 10.255.255.255
> 172.16.0.0 to 172.16.255.255
> 192.168.0.0 to 192.168.255.255
>With quality cards and proper ip addresses, your problems may just
>disappear.
No.
As long as he's is in his own Ethernet he can take
almost any IP he wants.
using "military" (normally no routed) IPs for LANs once were
the "cheapest firewall"...
But I miss him talking about default gateway, route etc...
To see if the card is "living" he should use
netstat -i
and watch the octets increasing...(or not)
If on the other end of the ethernet is a just booting, fresh setup Windows
there should be enough broadcasts, on ethernet level too.
>>I set it up as IBM NE2000 compatible - the i/o address is
>>300 and i/vector is 3 with no * to indicate a conflict.
Hm, Int 3 is usually serial port 2
The best(?) choice for that ISA crap was Int 10.
(I hope that the board is 16 bit..)
If not, give Int 5 a try.
To check hardware it is the easiest way to boot a linux CD
with a "live filesystem" (for example knoppix, which has a very
good hardware detection and mostly plays "out of the box" (without
any change to the system!)).
Tony Lawrence
> Tony Lawrence <to...@pcunix.com> wrote in message
> news:<ogVO9.525954$NH2.35331@sccrnsc01>...
>
>> Dan Skinner wrote:
>>
>>> Tom Parsons <c...@tegan.com> wrote in message
>>> news:<2002122617...@tegan.com>...
>>>
>>>
>>>> Carlo enscribed: | Hi
>>>
>>> <snip>
>>
>> Excuse me?
>>
>> BTW, it's predicament, but the advice I gave him after commenting
>> on the card was no different than what you gave him.
>
>
> Which goes to prove that you are not without redeeming value.
Right. But doesn't explain why you said that Tom and I offered nothing
toward fixing the man's problem.
>
>
>>> I agree that you have not selected top line stuff, but i assume
>>> your planned use is limited. You have selected in-appropriate
>>> ip's but since you connected them with a crossover cable the
>>> chance of internet damage is limited.
>>
>> That's not the point. The reason Tom and I mentioned that is that
>> this person will go on to other places and continue to use
>> inappropriate ip addresses. Also, just the posting of those
>> addresses can cause other people to copy them for their own
>> networks, ultimately causing more trouble for people like us who
>> clean up messes like that regularly.
>
>
> Then the goal was to educate and inform. The condescending attitude
> had me fooled.
(Sigh). Another person who thinks that everyone needs to be treated
with kid gloves. Good for you.
> I assume you do it for a living. Without the errors of beginners,
> you would have less work (Read: less income.)
You make no sense. Both Tom and I were trying to CORRECT the errors of
this beginner. It has nothing to do with income for either of us; it
was instead an atttempt to avoid the spread of more problems in the real
world. Do you understand that now?
(stuff deleted)
Somewhere on the net there is an interesting writeup of the problems
that these cheap cards have creating valid packets. I thought I had it
on my website, but I can't seem to find it (it may in fact be there, but
if so, I didn't index it well). The bits don't "get" confused - they
weren't right to start with if my memory serves me. Jeff L. might
remember more details than I do; I think he might have been the one who
first brought that article to my attention.
>> There's nothing wrong with Linksys for a home setup. I use a
>> similar device of a different brand here in my own home. I even
>> have small clients with Linksys routers. I wouldn't use it in a
>> larger setup, but there's nothing wrong with it if properly
>> configured (like not leaving the default password which I see in so
>> many places).
>
>
> And Tony, you leave your Cisco routers installed with default
> settings?
Criminy. Did you take a special course on pissing people off by not
reading what they wrote?
Let me say it again: THERE'S NOTHING WRONG WITH LINKSYS ROUTERS.
Did you hear me that time, or do you want to go on pretending that I
just automatically dislike everything you have to say?
Earlier in this post you said:
> I learned many things at IBM instructor training in the 60's . Two of
> them are: 1:( You'll motivate more with honey than with vinegar) 2:(
> If the student knows more than you do, call him sir and listen.)
It may be a long time before too many of us here are
addressing you in that manner. But don't feel that I think your posts
are worthless: you've done some good, helpful posts both here and in
some other groups I read. Unfortunately, you don't seem to like being
corrected, and don't read what other people say very carefully.
Keep up the good work, but try to pay closer attention.
Tony Lawrence
> (Tom Parsons) 26.12.02 in /comp/unix/sco/misc:
>>Step #2: Are you really "names.wjp.net"? If not, don't hijack their
>> ip address(es).
>
>
>> These blocks of IP addresses are reserved for private,
>>non-internet connnected networks. Use them.
>> 10.0.0.0 to 10.255.255.255
>> 172.16.0.0 to 172.16.255.255
>> 192.168.0.0 to 192.168.255.255
>>With quality cards and proper ip addresses, your problems may just
>>disappear.
>
>
> No.
> As long as he's is in his own Ethernet he can take
> almost any IP he wants.
> using "military" (normally no routed) IPs for LANs once were
> the "cheapest firewall"...
Not the point.
We wish to avoid the proliferation of these addresses because examples
encourage their use in places that WILL be connected.
Four or five times each year I have to re-number a network that some
inexperienced person had set up using real ip addresses. Sometimes the
switch is easy, sometimes it's a major piece of work, and sometimes it
is so horrible to contemplate that they just live with being unable to
reach certain parts of the internet.
Bill Vermillion
Dan Skinner <JDanS...@JDanSkinner.com> wrote:
>Tom Parsons <c...@tegan.com> wrote in message news:<2002122617...@tegan.com>...
>> Carlo enscribed:
>> | Hi
><snip>
>Carlo;
>Tony and Tom seem more interested in commenting on the quality of
>digital
>equipment, than on trying to help you out of the perdickerment.
They did try to help out of the 'perdickerment' by telling him to
get rid of the RealTek card. You just don't want to use it in a
multi-user envirionment.
From a recent post in annother NG.
========================================
> I have a P233, 128MB, running freebsd 4.6. It has a cheap realtek card.
The realtek card is poorly designed, so in order to get maximum
bandwidth usage from it you need a lot of CPU. From the comments
in src/sys/pci/if_rl.c:
"It's impossible given this rotten design to really achieve decent
performance at 100Mbps, unless you happen to have a 400Mhz PII or some
equally overmuscled CPU to drive it."
I recomend reading the rest of the comments at the beginning of
that file, as well as the bugs section of the rl(4) manpage.
....
Switch out the NIC or get a faster CPU. ...
========================================
And another post
========================================
>> I recomend reading the rest of the comments at the beginning
>> of that file, as well as the bugs section of the rl(4)
>> manpage.
>I have read these comments many times over the past year or
>two. I've swapped in an ed nic, and am receiving exactly the
>same results (not that ed is a particularily exciting chipset
>either.)
i had a realtek card running in a ftp server running FreeBSD and when
someone uploaded with full speed (100Mbit/s) top displayed 95% interrupt
and 5% cpu (roughly).
it really is the cheap rtl-chip which hinders performance. try an intel
or 3com card and see the difference!
========================================
Look at those figures and you will see why that the card is not
really a problem in a synchronous OS [eg MS products] as it get to
use the whole CPU if it want's. It's a pure pefromance killer in
any multi-tasking OS.
And if that isn't enough here is one paragraph from driver
source code from a Unix like OS
========================================
* The RealTek 8139 PCI NIC redefines the meaning of 'low end.' This is
* probably the worst PCI ethernet controller ever made, with the possible
* exception of the FEAST chip made by SMC. The 8139 supports bus-master
* DMA, but it has a terrible interface that nullifies any performance
* gains that bus-master DMA usually offers.
*
========================================
>I agree that you have not selected top line stuff,
There is quite a difference between 'top line stuff' and a chip
set that "redines the meaning of 'low end'".
>If they are new and you can return them, I would. I find that it
>usually pays to use SCO certified parts in SCO box.
Almost any good NIC card works. If you know your hardware and what
chips are on the card even a $20 card works. The $20 card in my
OSR5 machine that said IBM on the box thinks it's an Intel 100B.
Old Netgear cards had the DEC tulip chip while the current ones use
a 'lite-on' clone. That old chip was the chip iNTEL bought when
DEC sold its chip side and is the heart of many good NIC designs.
So being able to tell 'top-of-the-line' from 'the worst crap on the
planet' really has nothing to do with price if you know the HW
underlying the product. Most don't so you go with name brands.
>They are not really better,
But they ARE really better.
>but they are known to work.
There is a world of difference between 'working' and 'working well'
>The windows end doesn't
>matter as much, but while you are buying one, buy two. While I'm
>spending your money, buy a little, cheap switch, built into a
>cable/dsl router if you are looking forward to broadband. (Linksys so
>Tony and Tom can flame me) I have an SCO 3.2v4.2 desktop, an SCO
>3.2v5.0.4 server, and a Linux 2.4.8-26mdk
Why should you get flamed for Linksys? They work well if they
satfisfy the requirement you need.
The people [Tom, Tony and 100's of others on this group] who
maintain among them 1000s of machines certainly have more
real-world experience that I'd guess you have based on your
comment of 'SCO 3.2v4.2, OSR5.0.4 and a Linux system.
Bill
--
Bill Vermillion - bv @ wjv . com
Tony Lawrence
> The realtek card is poorly designed, so in order to get maximum
> bandwidth usage from it you need a lot of CPU. From the comments in
> src/sys/pci/if_rl.c:
>
> "It's impossible given this rotten design to really achieve decent
> performance at 100Mbps, unless you happen to have a 400Mhz PII or
> some equally overmuscled CPU to drive it."
>
> I recomend reading the rest of the comments at the beginning of that
> file, as well as the bugs section of the rl(4) manpage.
Thanks Bill: I couldn't remember where I had read this, but yes it's
"man 4 rl" on any convenient FreeBsd box :-)
The problems aren't actually what my memory said they were, but just as
bad or worse.
I'm not sure exactly what causes these to sometimes be unable to pass a
router, but I have seen them do this more than once. I don't think it's
a driver issue per se because I've seen it where one card installed
with the same drivers works but another does not. The observed symptoms
are that it sees the local network fine but just can't get "out". I
thought at first it might be this:
> When an interrupt is posted to signal that
> a frame has been received, it is possible that another frame might be in
> the process of being copied into the receive buffer while the driver is
> busy handling the first one. If the driver manages to finish processing
> the first frame before the chip is done DMAing the rest of the next
> frame, the driver may attempt to process the next frame in the buffer
> before the chip has had a chance to finish DMAing all of it.
but I would expect that to be a problem on low-end cpu's, and yet I've
seen this on 800 Mhz machines. Who knows? As the page also says:
> The RealTek data sheets are of especially poor quality: the grammar and
> spelling are awful and there is a lot of information missing, particu-
> larly concerning the receiver operation
The cards are just trouble. I rip 'em out at the first excuse.
(stuff deleted)
>
> Why should you get flamed for Linksys?
Of more interest is WHERE did he get flamed for Linksys? Not here,
though he seems to think otherwise.
Carlo
1)Prior to loading the one machine with Openserver I had Windows 98 on
the machine with the same network cards and crossover cable and they
networked fine.
2)The two machines are not on the internet which therefore does not
matter what IP addresses I use.
3) Sorry some of us are so ignorant but I'm sure that none of you
genius's woke up one day with all the answers and if you think you
have all the answers currently then you have a problem. The object of
these groups are for people to learn.
Thanks
Carlob
JDanS...@JDanSkinner.com (Dan Skinner) wrote in message news:<8ce22d01.02122...@posting.google.com>...
Bill Vermillion
A few years ago - when switches were about $100/port - one of the
high-end switch manufacturers was selling huge amounts of their
switch to customers who had used wrong addresses.
It was basically a double NAT/PAT devices.
[bad-ip:port] <-> [public-ip:port] <-> [correct-ip:port] <-> [the-world]
Eliminated renumbering of hundreds/thousands of machines, so you
could do a slow/planned migration. The only IPs which weren't
reachable from inside were those on the same IP outside.
Where there's a will there's a way if there's enough money.
Tony Lawrence
> Thanks guys for the advice but just to clear up some issues:
> 1)Prior to loading the one machine with Openserver I had Windows 98 on
> the machine with the same network cards and crossover cable and they
> networked fine.
Would have been better if you had told us that earlier, but it really
doesn't change much.
> 2)The two machines are not on the internet which therefore does not
> matter what IP addresses I use.
Not the point. You must have read our multiple explanations of this.
What part don't you understand?
> 3) Sorry some of us are so ignorant but I'm sure that none of you
> genius's woke up one day with all the answers and if you think you
> have all the answers currently then you have a problem. The object of
> these groups are for people to learn.
Oh for crying out loud. Another person who can't read and thinks he's
been insulted.
Yes, we allknow what the purpose of these groups is. Learning doesn't
have to be sugarcoated with "You are such a wonderful person but you've
just made the teeniest mistake with regard to your use of ip addresses
that belong to some one else. You probably don't understand why some of
us think that's important, but.."
As to thinking we have all the answers, few of us think that at all.
Most of us are quite sure that we are basically hopeless morons who have
had the goood fortune to pick up a little knowledge here and there and
we repay those who educated us by posting here. See
http://aplawrence.com/wiz.html for what I think of my own skills.
Bill Campbell
>Bill Vermillion wrote:
>
>> The realtek card is poorly designed, so in order to get maximum
>> bandwidth usage from it you need a lot of CPU. From the comments in
>> src/sys/pci/if_rl.c:
>>
>> "It's impossible given this rotten design to really achieve decent
>> performance at 100Mbps, unless you happen to have a 400Mhz PII or
>> some equally overmuscled CPU to drive it."
>>
>> I recomend reading the rest of the comments at the beginning of that
>> file, as well as the bugs section of the rl(4) manpage.
>
>Thanks Bill: I couldn't remember where I had read this, but yes it's
>"man 4 rl" on any convenient FreeBsd box :-)
>
>The problems aren't actually what my memory said they were, but just as
>bad or worse.
I had quite a bit of trouble with these cards about four years ago. If I
remember correctly we were using Compex PCI NE2000 clones which worked
reasonably well on Linux boxes -- until they changed to the Realtek
chipsets. I didn't remember the details, but only that they were trouble
and to be avoided.
....
>>
>> Why should you get flamed for Linksys?
>
>Of more interest is WHERE did he get flamed for Linksys? Not here,
>though he seems to think otherwise.
LinkSys had a reputation on Linux for only working properly on their
patched versions of the Tulip drivers.
I had serious problems with tulip drivers on a mail server several months
ago. We were getting huge numbers of errors on the NICs, with LinkSys and
Kensington NICs using the tulip drivers. I installed a 3c905 NIC, and the
errors dropped to zero (Caldera 3.1.1 Workstation, 2.4.13 kernel).
We generally keep a few 3c509 ISA and 3c905 PCI cards on hand for folks who
want to do things on the cheap. These are available used at RE*PC, 3COM
has a lifetime warranty on them which they honor, and they seem to work
reasonably well on pretty much anything, certainly better than crap like
the Realtek NICs. They're also well supported by all versions of SCO
OpenServer, and I've frequently had problems with newer cards.
Bill
--
INTERNET: bi...@Celestial.COM Bill Campbell; Celestial Software LLC
UUCP: camco!bill PO Box 820; 6641 E. Mercer Way
FAX: (206) 232-9186 Mercer Island, WA 98040-0820; (206) 236-1676
URL: http://www.celestial.com/
``Capitalism works primarily because most of the ways that a company can be
scum end up being extremely bad for business when there's working
competition.'' -rra
John Schmidt
On 27 Dec 2002, Carlo wrote:
> The object of these groups are for people to learn.
No, that's not the object. As "nobull" put it so succinctly in
comp.lang.perl.misc:
"Get real! This is a discussion group, not a helpdesk. You post
something, we discuss its implications. If the discussion happens to
answer a question you've asked, that's incidental."
JS
Jean-Pierre Radley
| Thanks guys for the advice but just to clear up some issues:
| 1)Prior to loading the one machine with Openserver I had Windows 98 on
| the machine with the same network cards and crossover cable and they
| networked fine.
Which only proves that the cards work under Windows and not under Unix.
| 2)The two machines are not on the internet which therefore does not
| matter what IP addresses I use.
It matters enough that Tom Parsons and I ran a two year campaign to
get SCO to replace all examples in all OSR 5 documentation that used
anything except canonically non-routable IP addresses, because some
people are in the habit of blindly copying examples.
| 3) Sorry some of us are so ignorant but I'm sure that none of you
| genius's woke up one day with all the answers and if you think you
| have all the answers currently then you have a problem. The object of
| these groups are for people to learn.
Part of a genius's genius is not to claim to be one of the geniuses.
But if you claim that those refuting your statements might be geniuses,
fine; just do a little learning from them yourself. So far, you're
stubbornly closing your ears.
--
JP
Bill Vermillion
Bill Campbell <bi...@celestial.com> wrote:
>On Fri, Dec 27, 2002 at 06:00:45PM +0000, Tony Lawrence wrote:
>>Bill Vermillion wrote:
>>> The realtek card is poorly designed, so in order to get maximum
>>> bandwidth usage from it you need a lot of CPU. From the comments in
>>> src/sys/pci/if_rl.c:
>>> "It's impossible given this rotten design to really achieve decent
>>> performance at 100Mbps, unless you happen to have a 400Mhz PII or
>>> some equally overmuscled CPU to drive it."
>>>
>>> I recomend reading the rest of the comments at the beginning of that
>>> file, as well as the bugs section of the rl(4) manpage.
>>Thanks Bill: I couldn't remember where I had read this, but yes it's
>>"man 4 rl" on any convenient FreeBsd box :-)
The 'man 4 rl' is far kinder than the comments in the source code
for the drivers.
>>The problems aren't actually what my memory said they were, but
>>just as bad or worse.
Those could be in the driver sources.
>>>
>>> Why should you get flamed for Linksys?
>>Of more interest is WHERE did he get flamed for Linksys? Not here,
>>though he seems to think otherwise.
>LinkSys had a reputation on Linux for only working properly on their
>patched versions of the Tulip drivers.
The reference I see on the sources I have here on Linksys and tulip
reference the windbond chipset which is almost like the 'tulip'.
The other tulip clone is a VIA and the source notes also reference
the DMA problems.
Since only DEC and now iNTEL made/make real 'tulip' chips unless
they use those they are using the clones.
>I installed a 3c905 NIC, and the errors dropped to zero (Caldera
>3.1.1 Workstation, 2.4.13 kernel).
>We generally keep a few 3c509 ISA and 3c905 PCI cards on hand
>for folks who want to do things on the cheap.
The olc 3c509B ISA cards just seem to work and work - though they
are no speed demon and running an ISA card in today's world surely
wastes machine resources.
Rainer Zocholl
>A few years ago - when switches were about $100/port - one of the
>high-end switch manufacturers was selling huge amounts of their
>switch to customers who had used wrong addresses.
>It was basically a double NAT/PAT devices.
>[bad-ip:port] <-> [public-ip:port] <-> [correct-ip:port] <->
>[the-world]
That reminds me to the "Most Dangerous Things In The World"
One was
"A software with a hardware patch."
another:
"A softworker with a solder iron."... (IIRC)
>Eliminated renumbering of hundreds/thousands of machines, so you
>could do a slow/planned migration. The only IPs which weren't
>reachable from inside were those on the same IP outside.
And if that's ".mil" then there is usually no real problem...
But of cause:
That's all history and it is good practise to recommend _strongly_
to use RfC IP in private network (tho that does not solve the
OPs problem ;-))
Bill Vermillion
Well I certainly learned a lot on the 'net when I first got news
access but that was back in the days when 90% of the posts were
helpful a figure which is inverted in many of today's NGs. But
with only 6MB of messages/day compared to the 300+GB the small
quanity was more than made up for by quality.
Bela - always the compulsive helper - was keeping things lively
over on the Dr Dobb's forum on Compuserve. BIX had a strong
technical emphasis. And event the alt.sources on usenet not filled
with effluvium we see today.
But it was more of a two-way street - you got help from someone and
you were expected to help others in your areas of expertise.
In those days the 'tech test' of just getting connected usually
indicated you had some clue as to how the world of computers
worked.
Now that the ability of getting connected consists almost solely of
being able to plug in power cords for the monitor and computer and
one for telephone any idiot can post, and sorry to say a great many
do.
))))
))))
:::: ))))
:::: ))))
---- ))))
---- ))))
:::: ))))
:::: ))))
))))
))))
Tony Lawrence
> Part of a genius's genius is not to claim to be one of the geniuses.
Hah. I can *prove* that I'm a moron. No wait, that would make me a
genius. OK, I can prove I'm a genius. No, that doesn't work either..
Darn. I know: I'm a curmudgeon. OP say so, so it must be true.
Seriously, one of the few things I have learned about intelligence is this:
If you observe those obviously less intelligent than yourself, you can
rank them pretty well: a is more clever than b, but less so than c, and
d is pretty close to me. You get that just from their conversation; you
see where they start to lose track, how far they can follow arguments,
how well they advance their arguments, and even by what their arguments are.
But when you observe those who are more intelligent, that's pretty much
all you know. You are not able to rank them very well at all. You
aren't smart enough to follow their arguments well, so you can't judge
their relative merit: all you know is that it is beyond your ability to
keep up. And you may even be too dumb (relatively) to even know that!
So whenever I find myself making a mental judgement of someone that puts
them rather far down the evolutionary scale, I try to remember that
there are probably lots of people with an equally disdainful view of me,
but I have no way of knowing. I may think I'm not that far away, and in
fact be a long way below. It doesn't change my opinion of the dolt, of
course, it just reminds me that it's all relative and we all can only
use what we have. The worst of us make better conversation than a dog.
Bob Meyers
"Bill Vermillion" <b...@wjv.comREMOVE> wrote in message
news:H7sn2...@wjv.com...
> The olc 3c509B ISA cards just seem to work and work - though they
> are no speed demon and running an ISA card in today's world surely
Well now, I have developed a sick hatred of 3Com cards since when I used
them last, a couple of years ago, it was always a case of "throw away all
documentation and software that came with your 3Com card, then download
something that might work from their web".
I switched to Intel NICs and never looked back.
Rainer Zocholl
>Thanks guys for the advice but just to clear up some issues:
>1)Prior to loading the one machine with Openserver I had Windows 98 on
>the machine with the same network cards and crossover cable and they
>networked fine.
A working windows is not a (slightest) hint that the hardware is OK!
MS did a lot of work to make Windows run on the crappiest hardware!
Please download a linux life filesystem CD, burn it, boot it
and see what it says.
at least try the command
"netstat -i" to see of your card is working.
I'm quite sure that you have an interrupt conflict.
>2)The two machines are not on the internet which therefore does not
>matter what IP addresses I use.
But it is easier to change those 2 addresses now.
In some time there may come a third PC, again with the
illegal address, and then you wish to access the internet...
use
10.1.1.1 net mask 255.255.255.0 for your server or gateway
and
10.1.1.11 net mask 255.255.255.0 for your client.
>3) Sorry some of us are so ignorant but I'm sure that none of you
>genius's woke up one day with all the answers and if you think you
>have all the answers currently then you have a problem. The object of
>these groups are for people to learn.
To say it simple:
"usenet is not a right"
Rainer Zocholl
The only disadvantage:
There is (was?) no SCO OSR5 driver for the intel gigabit NICs...
Bob Meyers
"Rainer Zocholl" <UseNet-Pos...@zocki.toppoint.de> wrote in message
news:8ciUu...@zocki.toppoint.de...
> (Bob Meyers) 27.12.02 in /comp/unix/sco/misc:
>
>
> The only disadvantage:
> There is (was?) no SCO OSR5 driver for the intel gigabit NICs...
The highest speed barb wire can go is my bottle neck, so I'm not in the giga
class just yet.
Bill Vermillion
Rainer Zocholl <UseNet-Pos...@zocki.toppoint.de> wrote:
>(Bill Vermillion) 27.12.02 in /comp/unix/sco/misc:
>>A few years ago - when switches were about $100/port - one of the
>>high-end switch manufacturers was selling huge amounts of their
>>switch to customers who had used wrong addresses.
>>It was basically a double NAT/PAT devices.
>>[bad-ip:port] <-> [public-ip:port] <-> [correct-ip:port] <->
>>[the-world]
>That reminds me to the "Most Dangerous Things In The World"
The above actuall worked well complete with the ability to set
secure tunnels between machines on the same LAN. You could do such
things as totally isolate accounting from the rest of the system,
so they would act as internal firewalls too. These were Xylan
devices at over $5K for 48 ports. Some of the big purchasers were
telcos and large corporations. It was cheaper even at that price
than renumbering a 1000+ computers.
>One was
>"A software with a hardware patch."
>another:
>"A softworker with a solder iron."... (IIRC)
Or as William ?? Godbout - Godbout Computers - early 1980s - was
widely quoted as saying "Never trust a programmer with a
screwdriver". Only those of us who remember CP/M will probably
remeber his name. The usenet maps maintainter for the state of
Florida always used to rib me about that as I came out of an
engineering type background. I never did like RF because it could
bite you when you weren't looking.
>>Eliminated renumbering of hundreds/thousands of machines, so you
>>could do a slow/planned migration. The only IPs which weren't
>>reachable from inside were those on the same IP outside.
>And if that's ".mil" then there is usually no real problem...
>But of cause:
>That's all history and it is good practise to recommend _strongly_
>to use RfC IP in private network (tho that does not solve the
>OPs problem ;-))
Absolutely. I haven't had any entanglement with machines being
assigned IPs that didn't belong to them for many years. Worst
problems any
Brian K. White
Heck I do that with every thing I ever install on any OS.
I keep the cd's but I haven't opened one in ages.
I check for firmware updates for anything that can possibly have one
too. Even my external usr modem on my desktop gets a new firmware
flash once in a while, but I'll admit that is an extreme example and
an exception, they have a fancy windows application that automatically
checks for updates and loads them and it's just my own personal desk
so I can afford to be a little silly with it. If the modem stops
working after an update, I'm prepared to deal with it.
yes, installs go a little slower, but I think I don't have to re-visit
a box as often. I don't really know, because I'd have to stop for a
while to compare, and I can't make myself do that, and since hardware
changes so fast it wouldn't be a meaningful comparison anyways.
Brian K. White
Sometimes I just put it like this:
This stuff is complicated enough at it's best, and there are more than
enough ways for problems to occur even in the best cases, that there
is really no room for coloring outside the lines.
you can of course do anything you want, but if you are going to use
unrecommended hardware, and unrecommended configurations, then few
others will be willing to help make it work because they will see it
as a waste of time & effort.
If your job is to carry water, but your pail has a leak. you are free
to spend all kinds of time and energy trying ways to plug the leak
like put tape over it, or glue it, or line the pail with a bag... but
all that stuff is going to be less effective and at the same time will
cost more in wasted time than simply getting a new good quality pail.
People who understand this will not waste their time doing something
pointless, and they won't waste their time helping you do it for the
same exact reasons.
I understand that sometimes the "new pail" is a several thousand $
peice of fancy hardware, and in cases like that it may be worth a
*little* time & effort to come up with a cheap work-around, but even
then, at most consultants billing rates, it has to be a pretty
expensive chunk of hardware and the cheap fix has to arrive in under a
day or two of fiddling before you are back in the same boat. But as
long as we are just talking about a nic and merely choosing certain
numbers over certain others to fill in a couple blanks... few here
would waste even 1/2 hour avoiding that.
And if after doing that you still have the problem, the expense was
*not* wasted. You've simply provided the minimum sane starting point
from which to start debugging and diagnosing. If doing that does
nothing else than act as the magic snake dance that gets one of us to
consider your problem worth their attention and the problem then gets
fixed, that alone makes it worth doing.
Mike Brown
>
> Carlo enscribed:
> | Hi
> | I loaded Openserver 5.04 on a machine at home. It has a Realtek
> | network card in it which is connected to another Realtek card in a
> | machine running XP Pro (crossover cable)
> | The IP address for the Unix machine is 199.1.1.10 (set up in
> | scoadmin).
> | I clear a line totally in VI?)
> | If I ping 199.1.1.10 it does not lose any packets. Is this a sure sign
> | that the network card is configured correctly - I set it up as IBM
> | NE2000 compatible - the i/o address is 300 and i/vector is 3 with no *
> | to indicate a conflict.
> | loading Openserver the machine was loaded with 98SE and the network
> | was up and fine - different IP address). If I ping 199.1.1.1 on the XP
> | machine it fails:
> | Network Adapters [00000009] Realtek RTL8139 Family PCI Fast Ethernet
> | NIC FAILED
> | Caption = [00000009] Realtek RTL8139 Family PCI Fast Ethernet NIC
> | DatabasePath = %SystemRoot%\System32\drivers\etc
> | Description = Realtek RTL8139 Family PCI Fast Ethernet NIC
> | DHCPEnabled = FALSE
> | DNSEnabledForWINSResolution = FALSE
> | DNSHostName = CARLO_XP
> | Index = 9
> | IPAddress = 199.1.1.1 (FAILED)
> | Pinging 199.1.1.1 with 32 bytes of data:
> | Request timed out.
> | Request timed out.
> | Request timed out.
> | Request timed out.
> | Ping statistics for 199.1.1.1:
> | Packets: Sent = 4, Received = 0, Lost = 4 (100% loss)
> | IPConnectionMetric = 20
> | IPEnabled = TRUE
> | IPFilterSecurityEnabled = FALSE
> | IPSecPermitIPProtocols = 0
> | IPSecPermitTCPPorts = 0
> | IPSecPermitUDPPorts = 0
> | IPSubnet = 255.255.255.0
> | IPXEnabled = FALSE
> | MACAddress = 00:50:AB:00:07:34
> | ServiceName = RTL8139
> | SettingID = {88F1AA4B-2319-40F9-A97B-9551DD22F5B2}
> | TcpipNetbiosOptions = 0
> | WINSEnableLMHostsLookup = TRUE
> |
> | and when I try ping 199.1.1.1 from the unix machine I get 'host is
> | down'.
> | Firstly, is the problem only on the XP side?
> | Is there anything else I need to set up on the Unix side to get the
> | network to talk. (I have an application on the Openserver side that I
> | need to basically telnet into from the XP side to run it)
> | I have been trying for some days now with no success. Please help.
>
> Step #1: Open case on each computer
> Remove the Realtek NIC
> Apply 16 lb sledge hammer to each card
> Deposit remains in appropriate receptable
> spent the difference in price fiddling with this junk.
> Install the better cards and go to Step #2
>
> Step #2: Are you really "names.wjp.net"? If not, don't hijack their
> ip address(es).
>
> These blocks of IP addresses are reserved for private, non-internet
> connnected networks. Use them.
> 10.0.0.0 to 10.255.255.255
> 172.16.0.0 to 172.16.255.255
> 192.168.0.0 to 192.168.255.255
>
> With quality cards and proper ip addresses, your problems may just disappear.
>
> ==========================================================================
> Tom Parsons t...@tegan.com
> ==========================================================================
Interesting discussion on NICS so far.
There are major performance differences between low end and high end, ie
entry level windows NICs and server NICS. A quick and dirty test to
run is the windows tcpspeed program from multiple workstations against
the UNIX server. The tcpspeed reads as fast as possible from the chargen
port, #19. Monitor the CPU usage under full load, and maximum network
throughput.
I have been particularly impressed with the Compaq/Intel 4 port 10/100
64bit 66MHz cards. Intel supplies a TransmitLoadBalancing and failover
driver that puts all 4 ports on the same IP. At full load with 8
workstations the throughput was close to 4 X 100M and the server was
under 4% CPU load ( Dual 1Ghz XEON ).
Completely different setup, but with a 'clone' board, single port 10/100
PCI that had a manufacturer supplied driver, a PII 350 CPU averaged
40% CPU.
If anyone gets the chance, download tcpspeed
( maximized.com/freeware/tcpspeed ) and post the results.
Mike
--
Michael Brown
The Kingsway Group
Thom Price
>
> These blocks of IP addresses are reserved for private, non-internet
> connnected networks. Use them.
> 10.0.0.0 to 10.255.255.255
> 172.16.0.0 to 172.16.255.255
> 192.168.0.0 to 192.168.255.255
>
You mean
172.16.0.0 to 172.31.255.255
no? Else it's the same number of IPs as the Class C private range...
HTH.
Thom
Thom Price
> Hi
> I loaded Openserver 5.04 on a machine at home. It has a Realtek
> network card in it which is connected to another Realtek card in a
> machine running XP Pro (crossover cable)
> The IP address for the Unix machine is 199.1.1.10 (set up in
> scoadmin).
> In etc/hosts there are two entries: 127.0.0.1 and 199.1.1.10. (How do
> I clear a line totally in VI?)
> If I ping 199.1.1.10 it does not lose any packets. Is this a sure sign
> that the network card is configured correctly - I set it up as IBM
> NE2000 compatible - the i/o address is 300 and i/vector is 3 with no *
> to indicate a conflict.
> I then set up the XP machine with an IP address of 199.1.1.1 (prior to
> loading Openserver the machine was loaded with 98SE and the network
> was up and fine - different IP address). If I ping 199.1.1.1 on the XP
> machine it fails:
> Firstly, is the problem only on the XP side?
> Is there anything else I need to set up on the Unix side to get the
> network to talk. (I have an application on the Openserver side that I
> need to basically telnet into from the XP side to run it)
> I have been trying for some days now with no success. Please help.
> Carlob
Yes, the problem is on the XP side. SCO side s/b OK, but heed
warnings on NICs.
I see it's Index 9, which means that it's your 9th attempt at loading
NIC drivers, I believe. Remove the Hidden Devices, and try again, or
reinstall from ground up. In W2K, you'd run Device Manager, then
click View, choose Show hidden devices. It'd allow you to remove all
traces of all NICs used and unused.
When you get the new NIC loaded, open the Properties for the NIC and
choose "Show icon in systray when connected" or something similar for
XP. (that's the W2K method, but XP s/b similar.) It will be an
indicator (not an absolute) that you do or don't have physical link.
You need to get XP ping to respond, before venturing further. On both
sides, successful pings to your own IP don't necessarily mean
completely correct NIC and driver setup, so don't use it as your only
check.
Try the NIC in a different slot in the PC - it may choose another IRQ,
which as mentioned elsewhere may be your problem.
At the risk of getting flogged, I'd say swap the NICs. If the SCO is
broken... bad NIC. Worth a quick look.
HTH.
Thom
Tom Parsons
yeah...and I've not even begun to celebrate New Years :-(
Thom Price
> Thom Price enscribed:
> | Tom Parsons <c...@tegan.com> wrote in message news:<2002122617...@tegan.com>...
> | > Carlo enscribed:
> | > | Hi
> | >
> | > These blocks of IP addresses are reserved for private, non-internet
> | > connnected networks. Use them.
> | > 10.0.0.0 to 10.255.255.255
> | > 172.16.0.0 to 172.16.255.255
> | > 192.168.0.0 to 192.168.255.255
> | >
> |
> | You mean
> | 172.16.0.0 to 172.31.255.255
> | no? Else it's the same number of IPs as the Class C private range...
>
> yeah...and I've not even begun to celebrate New Years :-(
glad to clarify... gladder to *join* ya!!
Have a very Happy, Healthy, Prosperous New Year, all.
Rob S
-
- These blocks of IP addresses are reserved for private, non-internet
- connnected networks. Use them.
- 10.0.0.0 to 10.255.255.255
- 172.16.0.0 to 172.16.255.255
- 192.168.0.0 to 192.168.255.255
Just going back over some old posts, and it occured to me that I've never seen
anyone use the 172.16 range on a private network. Why is this?
-Rob
robatwork at mail dot com
Ian Wilson
10.0.0.0 is easy to remember
192.168.0/24 is what cheap router manufacturers use for SOHO NAT routers.
or
Lazy people stop at the first thing they see.
People who care about IP-addressing pick small IP subnets for small
physical networks (hence frequent use of 192.168.x/24)
or
The bit patterns in 172.16. match a resonant frequency in 100-base-T and
cause sysadmins heads to explode when placed too near ethernet switches.
Take your pick :-)
--
Ian Wilson.
Jean-Pierre Radley
I've never seen snow in Norway.
A little further investigation would reveal that I've never been in
Norway.
Clear enough?
--
JP
Jeff Liebermann
<robatworkDelet...@mail.com> wrote:
>Just going back over some old posts, and it occured to me that I've never seen
>anyone use the 172.16 range on a private network. Why is this?
Fire up Google and inscribe "172.16." into the search box.
178,000 hits. One would suspect that using this IP address range is
sufficiently popular.
There at least two cheap dialup routers on the market that default to
an IP address in this range. I have one in my office but can't recall
the name.
Cisco uses all 3ea IP ranges in their IOS config examples.
Earthlink's UHP modem dialer defaults to 172.16.0.254
The original version of DirectTVDSL's DSL modem defaulted to
172.16.5.254.
I use it for internal administrative VLAN IP's at an ISP.
--
Jeff Liebermann 150 Felker St #D Santa Cruz CA 95060
(831)421-6491 pgr (831)336-2558 home
http://www.LearnByDestroying.com AE6KS
je...@comix.santa-cruz.ca.us je...@cruzio.com
Bill Vermillion
Ian Wilson <scob...@infotop.co.uk> wrote:
>Rob S wrote:
>> On Thu, 26 Dec 2002 23:22:31 GMT, Tom Parsons <c...@tegan.com> wrote:
>> -
>> - These blocks of IP addresses are reserved for private, non-internet
>> - connnected networks. Use them.
>> - 10.0.0.0 to 10.255.255.255
>> - 172.16.0.0 to 172.16.255.255
>> - 192.168.0.0 to 192.168.255.255
>>
>> Just going back over some old posts, and it occured to me that I've never seen
>> anyone use the 172.16 range on a private network. Why is this?
>
>10.0.0.0 is easy to remember
>
>192.168.0/24 is what cheap router manufacturers use for SOHO NAT routers.
>
>or
>
>Lazy people stop at the first thing they see.
>People who care about IP-addressing pick small IP subnets for small
>physical networks (hence frequent use of 192.168.x/24)
You can easily run a 255.255.255.0 netmask on a 10.x.x.x network
for the same effect.
>or
>The bit patterns in 172.16. match a resonant frequency in
>100-base-T and cause sysadmins heads to explode when placed too
>near ethernet switches.
That sounds logical.
But I've seen some 'sysadmins' head explode [not literally] under
far less stress.
Rob S
<je...@comix.santa-cruz.ca.us> wrote:
-Fire up Google and inscribe "172.16." into the search box.
-178,000 hits. One would suspect that using this IP address range is
-sufficiently popular.
Using your example, which doesn't prove much anyway, suggests that 192.168 is
8.5 times as popular and 10.0 is 24 times as popular.
I think JPR hit the nail on the head - all the private networks in Norway use
the 172.16 subnet - their sysadmins have nonresonating heads.
Thanks for clearing that one up.
Jeff Liebermann
<scob...@infotop.co.uk> wrote:
>People who care about IP-addressing pick small IP subnets for small
>physical networks (hence frequent use of 192.168.x/24)
Before you trivialize the problem, may I suggest you consider a
problem I have with such networks. If you build a router terminated
VPN, and BOTH ends use the same Class-C block, there will be problems.
All the docs and books warn against using the same Class-C block on
both sides, but it happens far too often.
A clueless expert buys a pile of routers with hardware VPN and sets
them up using the default IP block. Usually, that's 192.168.0.xxx.
He connects them to the internet and everything is working just fine.
Then, he tries to setup a VPN to the other similarly configured
offices. Instant duplicated IP addresses and routeing horrors. Some
VPN firewalls actually handle such situations (i.e. Sonicwall).
Others are just fatally confused.
>The bit patterns in 172.16. match a resonant frequency in 100-base-T and
>cause sysadmins heads to explode when placed too near ethernet switches.
I only use numerology and astrology only when desperate. I've used
172.16. and my head has not exploded. However, I've noticed that my
hat doesn't quite fit any more. Oh-oh.
Bill Vermillion
Jeff Liebermann <je...@comix.santa-cruz.ca.us> wrote:
>On Tue, 13 May 2003 15:16:09 +0000 (UTC), Ian Wilson
><scob...@infotop.co.uk> wrote:
>>People who care about IP-addressing pick small IP subnets for small
>>physical networks (hence frequent use of 192.168.x/24)
>Before you trivialize the problem, may I suggest you consider a
>problem I have with such networks. If you build a router terminated
>VPN, and BOTH ends use the same Class-C block, there will be problems.
>All the docs and books warn against using the same Class-C block on
>both sides, but it happens far too often.
And doesn't that come because MS uses the 'natural' netmask - eg
255.255.255.0 instead of the netmask you designate.
I first came across this when I had to alias base Cisco address
with Macs on subnets as the Mac can't have a gateway outside it's
block. While you can put a small MS IP block up at the top of the
range and still use a bottom IP gateway - eg 192.168.30.252
with a 255.255.255.240 netmask - and use the 192.168.30.1 as a
gateway.
I have a VPN setup with no problems spliting a 192.168.0.x
into two pieces with a 255.255.255.128 netmask. Cable on one end,
DSL on the other and only the cable side has fixed IP. Works
wonderfully well.
>A clueless expert buys a pile of routers with hardware VPN and sets
>them up using the default IP block. Usually, that's 192.168.0.xxx.
And using the default 255.255.255.0 netmask - instead of changing
it. ? Lets all boycott MS for screwing this up.
Bill
Bill Andersen
Bill Vermillion wrote:
> And using the default 255.255.255.0 netmask - instead of changing
> it. ? Lets all boycott MS for screwing this up.
For those who can't decide: Hell, let's just boycott MS period.
> Bill
> --
> Bill Vermillion - bv @ wjv . com
>
For Bottomers: Hell, let's just boycott MS period.
Jeff Liebermann
wrote:
>In article <1er4cv4qsc931nv4f...@4ax.com>,
>Jeff Liebermann <je...@comix.santa-cruz.ca.us> wrote:
>>On Tue, 13 May 2003 15:16:09 +0000 (UTC), Ian Wilson
>><scob...@infotop.co.uk> wrote:
>
>>>People who care about IP-addressing pick small IP subnets for small
>>>physical networks (hence frequent use of 192.168.x/24)
>
>>Before you trivialize the problem, may I suggest you consider a
>>problem I have with such networks. If you build a router terminated
>>VPN, and BOTH ends use the same Class-C block, there will be problems.
>>All the docs and books warn against using the same Class-C block on
>>both sides, but it happens far too often.
>
>And doesn't that come because MS uses the 'natural' netmask - eg
>255.255.255.0 instead of the netmask you designate.
Not quite but that's the "solution". To make it work with identical
class C blocks on both sides of the VPN, one needs:
1. Different subnets on both sides
2. Broadcasts traverse over the VPN so that the Windoze browser
functions.
3. Non-overlapping DHCP ranges to avoid IP duplication.
4. Different IP's for the various routers.
I've actually done this two or three times and will NEVER do it again.
It works with two terminators on the VPN, but gets seriously confusing
with 3 or more. I have one system that currently has 5ea VPN routers
that started out this way. I was trying to avoid renumbering the
entire LAN. After a few screwups, I gave up, did the airplane thing,
and setup a different class C IP block for each LAN.
I wouldn't blame MS for the problem. All the vendors (except Netopia)
start out with a netmask of 255.255.255.0 and few warn of the
complications of using all the defaults in a VPN. Some are just plain
broken. The SafeNet Windoze VPN client from Sonicwall (5.13 and 8.0)
break badly if you have more than one remote LAN configured to use the
same non-routeable LAN IP block. Even if you disable the remote VPN
configuration for one remote VPN in an effort to make it work, the
buggy piece of junk software still fails. I have to use multiple
configuration files for each remote VPN to make this piece of junk
software work (at $75/seat).
To many, sub-net configuration is mysterious and confusing. I have to
use a cheat sheet to get it straight. It's not something I would want
to inflict on a total beginner or home user. Many of my customers are
buying hardware VPN routers so that they don't have to install VPN
client software on all their home PC's. Methinks sub-netting would be
too difficult for these users. Just getting them to *NOT* use the
same class C IP block as what's used in the home office is difficult
enough. I've lost count of how many times I've had to walk them
through that ordeal. I thank the gods and the IETF for DHCP which
makes such changes relatively painless.
>I first came across this when I had to alias base Cisco address
>with Macs on subnets as the Mac can't have a gateway outside it's
>block. While you can put a small MS IP block up at the top of the
>range and still use a bottom IP gateway - eg 192.168.30.252
>with a 255.255.255.240 netmask - and use the 192.168.30.1 as a
>gateway.
>
>I have a VPN setup with no problems spliting a 192.168.0.x
>into two pieces with a 255.255.255.128 netmask. Cable on one end,
>DSL on the other and only the cable side has fixed IP. Works
>wonderfully well.
Good idea and that should work. Also note that almost all the
examples of router terminated VPN's on the Cisco TAS web pile use
radically different blocks of IP's at each LAN. The example I was
using has 10.0.xxx.xxx on one LAN, and 192.168.xxx.xxx on the other
end. I don't think it was just for clarity that they did this.
>>A clueless expert buys a pile of routers with hardware VPN and sets
>>them up using the default IP block. Usually, that's 192.168.0.xxx.
>
>And using the default 255.255.255.0 netmask - instead of changing
>it. ? Lets all boycott MS for screwing this up.
You'll also have to boycott all the VPN terminating router
manufacturers. Linksys, Netgear, DLink, Sonicwall, Netscreen, ad
nausium. They all default to 255.255.255.0. So does SCO on all their
LAN configs.
Bill Vermillion
Jeff Liebermann <je...@comix.santa-cruz.ca.us> wrote:
>On Wed, 14 May 2003 18:27:30 GMT, b...@wjv.comREMOVE (Bill Vermillion)
>wrote:
>>>A clueless expert buys a pile of routers with hardware VPN and sets
>>>them up using the default IP block. Usually, that's 192.168.0.xxx.
>>And using the default 255.255.255.0 netmask - instead of changing
>>it. ? Lets all boycott MS for screwing this up.
>You'll also have to boycott all the VPN terminating router
>manufacturers. Linksys, Netgear, DLink, Sonicwall, Netscreen, ad
>nausium. They all default to 255.255.255.0. So does SCO on all their
>LAN configs.
It's not the default setup that's the problem - it's the way MS
does things - at least from what I've seen in the past.
Define an address. 192.168.1.115 netmask 255.255.255.240
That is part of a 192.168.1.112 to 192.168.1.127 block
with usable IPs of .113 thru .126 and a broadcast of .127
In an MS machine you can set the gateway to be 192.168.1.1 and it
will work - and that is NOT supposed to be the way it works.
Your gateway is supposed to be in the net-block you have defined.
It's not the >default< netmask I'm talking about - but MS
assuming that you can address anything in the 'C' range of 255
address - even it if is outside of your network.
I ran across this when setting up show networks. These would be
mixed architectures and some would have routers on the show floor
demonstrating their wireless efforts and being able to access the
real world.
I could put an MS machine anywhere inside the block and use the .1
address on that particular Cisco ethernet link as a gateway.
But with routers, and non-MS machines, I've have to drop an alias
on top of the ethernet card that was inside the assigned net-block,
as the other devices could not see the gateway if it was outside
their net-block.
Those led to some interesting trouble shooting exercises at times.
Sometimes I'd have to route a T1 to the floor for the comm demos -
but with 32 serial ports on the router that was no problem - other
than the constant wiring changes.
Does this explanation of the problem make more sense with this
explanation?
Jeff Liebermann
wrote:
Oh, that problem. Been there. If the Windoze client is running a
static IP address (no DHCP), then assigning a default route outside of
the subnet mask doesn't work. However, use DHCP and it does work.
That's because the MS DHCP client has some kind of router discovery
protocol running. If it hears a router advertisement or broadcast,
and does not have a useable default route, then it just switches to
the broadcast IP for a default route. (MS can have multiple default
gateways, another abomination). If it already has a route to the
internet, it adds the new route to the router table (route /print) to
a maximum of 4ea gateways. The router discovery thing is part of the
same RFC that creates the 169.254.xxx.xxx default IP's that drive me
nuts on most DHCP configs. It's also an old security hole that's
still quite functional. See:
http://www.atstake.com/research/advisories/1999/rdp.txt
I can see where it would be a problem on a show network with many
routers advertising themselves as the best route to nowhere.
I dunno if this is still the case with the later Windoze incantations.
I'll try it in my office when I have time.
Bill Vermillion
Jeff Liebermann <je...@comix.santa-cruz.ca.us> wrote:
>On Thu, 15 May 2003 18:27:32 GMT, b...@wjv.comREMOVE (Bill Vermillion)
>wrote:
>>In article <6re7cv41mbj8n5i4h...@4ax.com>,
>>Jeff Liebermann <je...@comix.santa-cruz.ca.us> wrote:
>>>On Wed, 14 May 2003 18:27:30 GMT, b...@wjv.comREMOVE (Bill Vermillion)
>>>wrote:
>>>>And using the default 255.255.255.0 netmask - instead of changing
>>>>it. ? Lets all boycott MS for screwing this up.
>>>You'll also have to boycott all the VPN terminating router
>>>manufacturers. Linksys, Netgear, DLink, Sonicwall, Netscreen,
>>>ad nausium. They all default to 255.255.255.0. So does SCO on
>>>all their LAN configs.
>>It's not the default setup that's the problem - it's the way MS
>>does things - at least from what I've seen in the past.
>>Define an address. 192.168.1.115 netmask 255.255.255.240
>>That is part of a 192.168.1.112 to 192.168.1.127 block
>>with usable IPs of .113 thru .126 and a broadcast of .127
>>In an MS machine you can set the gateway to be 192.168.1.1 and it
>>will work - and that is NOT supposed to be the way it works.
>>Your gateway is supposed to be in the net-block you have defined.
....
>>Does this explanation of the problem make more sense with this
>>explanation?
>Oh, that problem. Been there. If the Windoze client is running a
>static IP address (no DHCP), then assigning a default route outside of
>the subnet mask doesn't work.
I've found the opposite - that if it's static and the netmask does
not include the gateway address in the block the MS machine will
still work. Haven't verified this on new ones but the old 95/98s
surely did it this way.
> However, use DHCP and it does work.
We were running pure static and no DHCP.
A funny aside - I was at a computer show once and there was an
announcement over the PA asking that everyone running MS machines
check to see they weren't running a DHCP server - as it was causing
problems for the provider also running DHCP. :-)