a very stange networking problem with a WiFi provider in France

Hello everyone,

Here it is, in a nutshell: this prepaid WiFi provider, Meteor
Networks, allows you to browse fine if you are using Windows yet you
get garbled IP transmissions under Linux - or at least so it seems.
Here's the summary of it:

http://technonutsnbolts.blogspot.com...-gates-or.html

Any help much appreciated.

Thanks.

Boris.

"borepstein@gmail.com" <borepstein@gmail.com> hath wroth:

>Here it is, in a nutshell: this prepaid WiFi provider, Meteor
>Networks, allows you to browse fine if you are using Windows yet you
>get garbled IP transmissions under Linux - or at least so it seems.
>Here's the summary of it:
>
>http://technonutsnbolts.blogspot.com...-gates-or.html

The first step to solving a problem is to blame someone. It seems
you're blaming the service provider for pandering to Microsoft. That
takes care of that problem, so we can move on to a solution.

The problem is that your 2nd hand description of the problem is devoid
of any equipment lists, software versions, and configuration
information. It doesn't even include which Windoze or Linux mutation
is being used. In short, there's no way to determine if there's
anything happening in hardware or software. That's because all 802.11
wireless is done on IP layer 2 (MAC layer). Since Linux handles this
layer quite differently than Windoze, a Linux configuration or setting
error would certainly cause this problem.

The good news is that since your friend didn't provide even the exact
error messages he was getting, I'm relieved of the responsibility of
providing a detailed and exact answer. Therefore, I can guess(tm). My
guess is that there's something screwed up in the Linux wireless
driver or MAC layer configuration. It could also be in the access
point. Perhaps the Meteor network is using flow control and his Linux
setup is ignoring the management packets. No way to tell from here.

Instead of fixing the problem, it would be most interesting to see if
the problem is unique to Meteor. As your friend to drag his laptop to
some other hotspot and see if the problem persists. Your article
doesn't indicate if he's done any testing at other hot spots using
Linux.

The "missing blocks of web pages" is fairly common with a weak signal
or in a high interference environment. The browser simply gives up
trying to load some external URL's on the page and instead supplies a
blank box. This is considered acceptable instead of the alternative,
which was formerly to sit forever and wait for some obscure external
URL to load. I suggest you interrogate your friend for signal
strength, signal to noise ratio (or noise level), and connection
speed.

Have him also run a continuous ping test to the wireless router and
look for wide variations in latency. It should be fairly low
(2-8msec) and fairly consistently the same value. If not, he has
packet loss, interference, a weak signal, or all the aformentioned.

If you have any furthur questions, please include the missing models,
versions, error messages, signal readings, and settings. Also, don't
forget to include the requisite "it's all Microsoft's fault" or "Linux
rules" in your reply.


--
Jeff Liebermann jeffl@cruzio.com
150 Felker St #D http://www.LearnByDestroying.com
Santa Cruz CA 95060 http://802.11junk.com
Skype: JeffLiebermann AE6KS 831-336-2558

Jeff Liebermann wrote:
> I can guess(tm).

Wasn't the trademarked by Radio Shack?

Oh, no...they used "You got questions? We got blank stares"

It was IBM that copyrighted "This page left intentionally blank".

Hi Jeff,

Thanks for your response, though I must say it is a rather strange
one. See my comments below.

Boris.

On Aug 1, 2:26 pm, Jeff Liebermann <je...@cruzio.com> wrote:

>
> The first step to solving a problem is to blame someone. It seems
> you're blaming the service provider for pandering to Microsoft. That
> takes care of that problem, so we can move on to a solution.

Well... are you projecting?

No, I am not seeking someone to blame when I encounter a problem. I
analyse the problem and then - if evidence points to any particular
factor as being the culprit - only then do I blame this factor. Often
- just as here - in a hypothetical way.

Now as for details, we have a dual boot SuSE Linux 10.1/Windows XP
laptop that here in Boston does networking just fine, both WiFi (at
the owner's home) and Ethernet when he brings it to the office under
both OS's. Then he goes to Paris and at the hotel his laptop connects
just fine under Windows but has severe problems accomplishing this
feat under Linux. Is it so wrong after all this to hypothethize that
Windows (or something related thereto) may be at fault here?

>
> The problem is that your 2nd hand description of the problem is devoid
> of any equipment lists, software versions, and configuration
> information. It doesn't even include which Windoze or Linux mutation
> is being used. In short, there's no way to determine if there's
> anything happening in hardware or software. That's because all 802.11
> wireless is done on IP layer 2 (MAC layer). Since Linux handles this
> layer quite differently than Windoze, a Linux configuration or setting
> error would certainly cause this problem.

Well, my description is certainly somewhat short on detail, that is
valid criticism. I am not 100% sure what the differences are in the
MAC layer transport handling between Windows and Linux but from my
extensive experience with Linux I must say I never experienced any
problems with the Linux implementation. And, like I said before, that
Linux box does WiFi just fine on the regular basis.

>
> The good news is that since your friend didn't provide even the exact
> error messages he was getting, I'm relieved of the responsibility of
> providing a detailed and exact answer. Therefore, I can guess(tm). My
> guess is that there's something screwed up in the Linux wireless
> driver or MAC layer configuration. It could also be in the access
> point. Perhaps the Meteor network is using flow control and his Linux
> setup is ignoring the management packets. No way to tell from here.

You were never under any obligation to start with. Participation in
newsgroups like this one is entirely voluntarily. And I do thank you
for taking the time to try and help me here:)

>
> Instead of fixing the problem, it would be most interesting to see if
> the problem is unique to Meteor. As your friend to drag his laptop to
> some other hotspot and see if the problem persists. Your article
> doesn't indicate if he's done any testing at other hot spots using
> Linux.

Like I said, he uses WiFi on the regular basis.

>
> The "missing blocks of web pages" is fairly common with a weak signal
> or in a high interference environment. The browser simply gives up
> trying to load some external URL's on the page and instead supplies a
> blank box. This is considered acceptable instead of the alternative,
> which was formerly to sit forever and wait for some obscure external
> URL to load. I suggest you interrogate your friend for signal
> strength, signal to noise ratio (or noise level), and connection
> speed.

Well, would the signal-related characteristics vary wildly depending
on what OS you have booted? I would think that unlikely.

>
> Have him also run a continuous ping test to the wireless router and
> look for wide variations in latency. It should be fairly low
> (2-8msec) and fairly consistently the same value. If not, he has
> packet loss, interference, a weak signal, or all the aformentioned.

Makes sense, we may do that should there be a need.

>
> If you have any furthur questions, please include the missing models,
> versions, error messages, signal readings, and settings. Also, don't
> forget to include the requisite "it's all Microsoft's fault" or "Linux
> rules" in your reply.

It is not about that - it's about fixing the problem. Now as for a
more realistic scenarion - someone has suggested to me on a different
discussion board that possibly Meteor scans machines for open ports
and tries to block those that have open ports which in their opinion
makes them a potential server. I think that is plausible - so I
emailed him asking him to turn the firewall on and see what happens.

We'll see.

>
> --
> Jeff Liebermann je...@cruzio.com
> 150 Felker St #D http://www.LearnByDestroying.com
> Santa Cruz CA 95060http://802.11junk.com
> Skype: JeffLiebermann AE6KS 831-336-2558

> so I
> emailed him asking him to turn the firewall on and see what happens.

So your help to someone using wifi didn't include using a firewall FROM THE
START?

Perhaps your consulting is questionable, at best.

On Aug 5, 10:14 pm, "Bill Kearney" <wkearne...@hotmail.com> wrote:
> > so I
> > emailed him asking him to turn the firewall on and see what happens.
>
> So your help to someone using wifi didn't include using a firewall FROM THE
> START?
>
> Perhaps your consulting is questionable, at best.

Perhaps, and I never positioned myself as a WiFI expert. Then again,
most of the time he does WiFi at home with WEP, and all that is open
on the machine is SSH (no Telnet, FTP or other unsafe protocols as
near as I remember). So this is not a glaring security issue.

On Aug 2, 10:05 am, "borepst...@gmail.com" <borepst...@gmail.com>
wrote:
> Thanks for your response, though I must say it is a rather strange
> one. See my comments below.

All my responses are strange. What I lack in diplomacy, I make
up in experience, technical expertise, and insults.

> > The first step to solving a problem is to blame someone. It seems
> > you're blaming the service provider for pandering to Microsoft. That
> > takes care of that problem, so we can move on to a solution.
>
> Well... are you projecting?

Nope. Kindly explain the title of your blog entry:
"Web-pandering to Mr Gates or..."
I'm certainly no fan of Microsoft business practices. However, the
title is not particularly relevant to any part of your blog entry.
It's
also misleading as the "or..." part implies there's more to the
connection between Mr Gates and your friends problems. Change
the title and I don't have a problem.

> No, I am not seeking someone to blame when I encounter a problem.

The blog title suggests otherwise. The title of your usenet news
article also suggest a culprit, but it less definitive in that it
doesn't
actually blame Mr Gates. Hint: It's only a "strange problem" until
you find the answer.

> Now as for details, we have a dual boot SuSE Linux 10.1/Windows XP
> laptop that here in Boston does networking just fine, both WiFi (at
> the owner's home) and Ethernet when he brings it to the office under
> both OS's. Then he goes to Paris and at the hotel his laptop connects
> just fine under Windows but has severe problems accomplishing this
> feat under Linux. Is it so wrong after all this to hypothethize that
> Windows (or something related thereto) may be at fault here?

Easy. You make the assumption that whatever he's trying to accomplish
in France should be identical in function to doing the same thing in
the USA. That's often not the case, as different systems block or
throttle specific ports. Authentication is sometimes a problem. The
number of open outgoing TCP sockets are sometimes restricted.
Lots of possibilities that need to be checked. I'll grant you that
things
should work identically in two different countries, but there's no
guarantee.

I had a similar adventure with a friend using XP SP2 in France. She
was using the MS Language bar and automatically changed to French
when she arrived in Paris. That screwed up all her email settings,
and automatically switched her to an old Earthlink network setup that
we thought had been vaporized. I spent an hour on the phone trying
to figure what had changed. Switching back to English solved most
of the problems (except for the timezone). The moral is to not make
assumptions, they're usually wrong.

> Well, my description is certainly somewhat short on detail, that is
> valid criticism. I am not 100% sure what the differences are in the
> MAC layer transport handling between Windows and Linux but from my
> extensive experience with Linux I must say I never experienced any
> problems with the Linux implementation. And, like I said before, that
> Linux box does WiFi just fine on the regular basis.

There are substantial difference between Linux and Windoze network
settings and implementation. I don't want to go into detail. I
offered
such differences in settings as a possible culprit because you can't
really assume that the Linux and Windoze drivers load the wireless
card with exactly the same settings. Many wireless cards get most
of their functionality from software, not firmware. If it were all in
the
firmware, I might suggest the two operating systems were identical.
However, they're not, and you should not make that assumption even
if he's using NDISWRAPPER. Each operating system has to be
tested and configured individually.

> You were never under any obligation to start with. Participation in
> newsgroups like this one is entirely voluntarily. And I do thank you
> for taking the time to try and help me here:)

I do it for amusement. You were also not under any obligation to
provide sufficient information to identify the source of the problem.
We can continue to discuss procedural and theoretical issues
if you prefer.

> > Instead of fixing the problem, it would be most interesting to see if
> > the problem is unique to Meteor. Ask your friend to drag his laptop to
> > some other hotspot and see if the problem persists. Your article
> > doesn't indicate if he's done any testing at other hot spots using
> > Linux.
>
> Like I said, he uses WiFi on the regular basis.

Could I trouble you to ask him the question exactly as I asked. I
would like to know if a different service provider works differently.
I have no interest in his regular habits. Again, let me point out
that you are making a fairly bad assumption.

> Well, would the signal-related characteristics vary wildly depending
> on what OS you have booted? I would think that unlikely.

Actually they do. I've often been mystified at the large changes
in indicated signal strength between different versions of Windoze
drivers. Intel 2200BG Proset drivers drove me nuts with this
problem for quite a while. In addition, they also juggled the
firmware, which added to the confusion. The problem is that
drivers sometimes estimate signal strength using the packet
error rate instead of the actual signal level. That results in some
rather bizarre indicated signal strengths. I really don't know if
the difference is that drastic between Linux and Windoze as I've
never needed to check. Again, it would a bad assumption to
assume that the card parameters are identical in both Linux and
Windoze.

You're also assuming that whatever driver is installed, works well
with whatever access point is being used in France. Wi-Fi
compatibility
is fairly good, but there have been occasional problems. I don't
recall the exact hardware, but I've seen access point where only
a few wireless clients can successfully connect. This is usually
a timing issue between the access point and wireless client and
often solved with a firmware or driver update. If there is a huge
packet loss or timing issue, the ping test should show it.

> > Have him also run a continuous ping test to the wireless router and
> > look for wide variations in latency. It should be fairly low
> > (2-8msec) and fairly consistently the same value. If not, he has
> > packet loss, interference, a weak signal, or all the aformentioned.
>
> Makes sense, we may do that should there be a need.

He should see consistent 2-5 msec ping times to the gateway
router. Any larger delays indicates packet loss and/or
retransmissions.
That could be poor signal strength in either direction, but can also
be interference. Make sure he runs ping for a few minutes as
the default 4 pings will not show much.

> It is not about that - it's about fixing the problem. Now as for a
> more realistic scenarion - someone has suggested to me on a different
> discussion board that possibly Meteor scans machines for open ports
> and tries to block those that have open ports which in their opinion
> makes them a potential server. I think that is plausible - so I
> emailed him asking him to turn the firewall on and see what happens.

I hadn't heard that one, but it's possible. I don't know of any
wireless
provider that offers intrusion detection services for their clients.
They
may scan their own ports but not those of their clients. To do so
would
simply take far too long and generate far too much traffic. It also
would
need to be initiated from the gateway router as such probes would not
be effective through a hotel firewall/router unless the hotel offers
routeable
IP's to it's wireless clients. However improbable, using the Windoze
and Linux firewall in a possibly hostile location might be a good
idea.
I would be less worried about attacks from the internet, than from
other
users on the wireless LAN.