Page Cannot be Diplayed Error

  • Thread starter Thread starter Page Cannot Be Displayed
  • Start date Start date
P

Page Cannot Be Displayed

Hello All,

We have a strange error and cannot seem to find a fix.
For some reason, certain computers with IE 6.0 come up
with a page cannot be displayed error. However, when you
telnet to the site or use Netscape that site comes up no
problem. We have compared IE settings on computers that
work with those that do not work. Everything is exactly
the same on both system, but the page loads on one and
not the other.

Has anyone found a solution to this problem?

THanks,

Jeff
 
Jeff,

We went thru the same issue. We are still going thru the QA pace now with
some changes to the web server that addressed this problem with IE for us,
but there still seems to be some problems yet addressed.

Here is my summary of my work:

Basically, what you and thousands of other IE users are experiencing is a
RST (socket reset) problem. This will cause your browser to do one or
more of the following:

A) Intermittent "Page Cannot Be Displayed" errors

B) Intermittently the URL request is resent 2 or 3 automatically by the
browser,

C) And possibly, a lost of Basic Authentication, if being used on the web
site.

With C, this is probably augmented with recent CROSS DOMAIN issues which is
specifically addressed in the latest IE patch MS03-048.

I think it is all related.

How do you get a RST?

Well it depends.

The short answer is that the web server response page being sent to the
browser is not making it all the way to the browser before it receives a
socket close request, hence a "out of sequence" and thus a RESET.

The HTTP RFC specifically talks about a browser resenting a request if it
receives a RESET, this this make explain behavior B.

Based on some user reports, if your have "Show Friendly HTTP responses," in
your browser, you can see behavior A or B. So basically behavior A and B
are the same issue.

The long answer:

1) I believe the either IE and/or the windows socket layer has changed in
recent updates in regards to how it supports the SO_LINGER socket option.
This doesn't seem to be working anymore or it depends on a few things as
discussed below.

2) The web server may have to be changed to use TCP Half Close methods.
This is documented in MSDN under the shutdown() command and in the bible
book "TCP/IP Illustrated Volume 1" (Page 238). This helps support a
"graceful socket disconnection" by informing the browser the web server is
about to close and to flush the receive bucket.

3) and related to #2, if you are working with POSTed data, the web server
has to check and make sure it is supporting a possible extra CRLF (carriage
return/linefeed) after the posted data sent to the web server.

4) However, even though #2 and #3 help solved the problems and in technical
view, make better designs to be used by developers, it doesn't why #1
doesn't work any more, hence....

5) In my view, I believe there is probably a "caching" or "network
transmission timing" issue which explains why some people see the problem
and others do not. For example, I can reproduce the problem over a network
with ADSL but I can't duplicate it at T1 or local machine testing, in
addition...

6) The problem seems to also go away if there is a FIREWALL or something in
between. In other words, while I can repeat the problem over a ADSL
connection, when the FIREWALL is up, it is not repeatable.

So I believe that something has changed in either IE SO_LINGER option or how
the Windows socket deals with packet transmissions. Maybe this is all
related to Microsoft attempts to secure the system with internal firewalls
and/or their IP Helper API that offers TCP/IP operations in Windows 2000 and
better operating systems.

Also, in regards to #6, to throw in another monkey wrench, I have read in
recently with many efforts by large ISPs in the area of "IP Analytics"
technology in the attempt to perform pattern recognition (for virus, fraud
detection, etc). Could it be that these middle man are screwing up the
TCP/IP packet timing sequence?

Who knows!?

All I know "it" has changed. We have had a Web Server in the market since
1996 and as a small company we feel every bump in the road in regards to IE.
Supporting customers is a big plus for us and issues like this can't be
ignored when they report "IE has problem with my Wildcat! Web server setup."
So even though we think the changes made to our web server were long coming
(TCP Half Close support and watching for POST data extra CRLF bytes) and has
been verified to fixed to the problem for the customers reporting problem,
we have never seen the issues that have come about in recent weeks with IE.
The latest issue is the behavior C and the lost of Basic Authentication.
This might be related to CROSS FRAMES though. MS03-048 does not fix the
problem.

Hope this info provides some insight.
 
I'm having the same problem, Hector. I don't understand
much of what you told Jeff. If I don't find a solution, I
guess my option is to pay Microsof support to help me.
Any other ideas? Thanks Jerry
 
From what I read in this fora, Microsoft has the following:

http://support.microsoft.com/default.aspx?scid=kb;en-us;818139

where you can call them and get a hotfix.

I guess they are still in QA (testing) with this and are not quite ready to
release it but will give it to people who need it, if it solves the problem
for them.

From what I read, it sounds they have not 100% completely figured out, but
it is an intermittent issue (doesn't happen to everyone and it all depends
on a few thing when it does).

In Layman terms, this is my I think after all my technical analysis:

When you connect to the internet, you have two speeds, a DOWNLOAD speed and
a UPLOAD speed. For example, for CABLE or ADSL people, you download much
faster than you can upload.

So when a browser talks to a web server:

"Hey, Give me this page on this web site!

The web server sends this page (called HTTP response).

What is basically happening is that while Web Browser is reading the
response, it is being told to close the communications channel, the Web
browser sees this and says one of two things depending on your setup and
connection speed:

"Hey, I didn't get everything. I am going to ask you again!
Give me this page!

or just tells you:

"Hey boss, I can't display this page you asked for, the
connection broke"

when it fact, it probably GET everything from the web server, hence lies the
problem. IE is some how broken in that it is "see the close" and reacting
to the close must faster than it is reading the web response.

So for some people, you see them reporting "Page Cannot be Displayed"
errors and for others, you see them reporting "My browser is automatically
refreshing, why?"

In a nut shell, that is basically what's going on.

People who use are POSTING data from via the Web browser to a web site, is
another way to show the problem which is what happen to us when a customer
reported:

"Hey, IE is saying Page Cannot Be displayed when we post X amount
of data from a form"

I would say the same problem is occurring with Web Servers using XML, SOAP,
Web Services (i.e, .NET) which is basically sending POSTED data to a web
server.

There is a known compatibility issue with Browsers and Servers where there
is an extra 2 bytes after the posted data. The standard is to have the
extra two bytes. Some web servers don't expect it, others are not reading
the extra bytes correctly.

So what is happening here with IE at least, is that if those 2 extra bytes
are NOT ready by the web server, IE will show one of two above behaviors of
Page Not found and/or Resending the Page Request.

So either IE added those two extra bytes to COMPLY with the standard thus
breaking many servers out there and/or they also broke at the same time how
to read web server responses.

Of course, I could be off base, but I don't think so. <g>

Hope this gives you some insight. Check out the KB article.
 
Thanks, Hector. That's helpful.
-----Original Message-----
From what I read in this fora, Microsoft has the following:
us;818139

where you can call them and get a hotfix.

I guess they are still in QA (testing) with this and are not quite ready to
release it but will give it to people who need it, if it solves the problem
for them.

From what I read, it sounds they have not 100% completely figured out, but
it is an intermittent issue (doesn't happen to everyone and it all depends
on a few thing when it does).

In Layman terms, this is my I think after all my technical analysis:

When you connect to the internet, you have two speeds, a DOWNLOAD speed and
a UPLOAD speed. For example, for CABLE or ADSL people, you download much
faster than you can upload.

So when a browser talks to a web server:

"Hey, Give me this page on this web site!

The web server sends this page (called HTTP response).

What is basically happening is that while Web Browser is reading the
response, it is being told to close the communications channel, the Web
browser sees this and says one of two things depending on your setup and
connection speed:

"Hey, I didn't get everything. I am going to ask you again!
Give me this page!

or just tells you:

"Hey boss, I can't display this page you asked for, the
connection broke"

when it fact, it probably GET everything from the web server, hence lies the
problem. IE is some how broken in that it is "see the close" and reacting
to the close must faster than it is reading the web response.

So for some people, you see them reporting "Page Cannot be Displayed"
errors and for others, you see them reporting "My browser is automatically
refreshing, why?"

In a nut shell, that is basically what's going on.

People who use are POSTING data from via the Web browser to a web site, is
another way to show the problem which is what happen to us when a customer
reported:

"Hey, IE is saying Page Cannot Be displayed when we post X amount
of data from a form"

I would say the same problem is occurring with Web Servers using XML, SOAP,
Web Services (i.e, .NET) which is basically sending POSTED data to a web
server.

There is a known compatibility issue with Browsers and Servers where there
is an extra 2 bytes after the posted data. The standard is to have the
extra two bytes. Some web servers don't expect it, others are not reading
the extra bytes correctly.

So what is happening here with IE at least, is that if those 2 extra bytes
are NOT ready by the web server, IE will show one of two above behaviors of
Page Not found and/or Resending the Page Request.

So either IE added those two extra bytes to COMPLY with the standard thus
breaking many servers out there and/or they also broke at the same time how
to read web server responses.

Of course, I could be off base, but I don't think so. <g>

Hope this gives you some insight. Check out the KB article.

--
Hector Santos
WINSERVER "Wildcat! Interactive Net Server"
support: http://www.winserver.com
sales: http://www.santronics.com




.
Click to expand...
 
Back
Top