Closed Bug 87962 Opened 23 years ago Closed 23 years ago

Link causes download as weird *cache-control* file, Mozilla 0.9.1

Categories

(Core :: Networking: HTTP, defect, P3)

Product:
Core
Component:
Networking: HTTP
Type:
defect

Tracking

()

Status:
VERIFIED WORKSFORME
Milestone:
mozilla1.0

People

(Reporter: vance, Assigned: darin.moz)

References

(
URL
)

Details

The link supplied is the login screen for TIAA CREF retirement accounts.  I 
suppose to debug this you'll need to find someone with such a login to this site, 
or perhaps this description will be enough to go by.  Platform: OpenVMS, Mozilla 
build 2001060807, I doubt this has any VMS aspects, it probably affects all 
platforms.

Log into this site and chose the "Show Accumulation" link.  Instead of getting a 
page describing one's retirement accounts, like Mozilla 0.9 did, instead you get 
a non resizable window "Enter name of file to save to..." which does not show all 
that is in this window.  Under Files of type: Mozilla gives "*.1 cache-control: 
private cache-control: max-age=0" and under the default File name to save, 
Mozilla gives: "0.21%20cache-control%3A%20private%20cache-control%3A%20max-age=0"
which I can't see all of but which I can force to see by clicking the save button 
when the file already exists.  

I suspect that Mozilla is getting mixed up on interpreting the content header 
with the cache-control header and assuming that the cache-control header is the 
type of file, and not recognizing this type is offering to download it.

Mozilla 0.9 did not have this problem with this link.  Other links on the site 
work fine.
I tried this using Netscape 6.1 Preview release 1 on a Mac and it has the same
problem with this link (which is actually "See Accumulation), i.e. it tries to
download the file using the weird cache-control type.  However, Netscape 4.77
on Mac displays this link ok and Page Info shows the Mime Type as "text/html".
OS: OpenVMS → All
Hardware: DEC → All
can one of you all attach the source or a simplified testcase that demonstrates
the problem.  Over to networking for now.
Assignee: asa → neeti
Status: UNCONFIRMED → NEW
Component: Browser-General → Networking
Ever confirmed: true
QA Contact: doronr → benc
need test account or a simplified test case to try this out
Keywords: qawanted
Priority: -- → P3
Target Milestone: --- → mozilla0.9.3
vance: could you try to generate a packet trace of the HTTP download?  there are
a lot of tools out there that can help (though i'm not sure what you could use on
OpenVMS)... there is a good tool at ngrep.sourceforge.net which may help.  at any
rate, if you can get a dump of the tcp/ip activity on port 80 that would allow us
to see exactly what the server is sending.  i suspect that we are having trouble
parsing some HTTP header.  thanks in advance!
I used MultiNet's TCPVIEW program to examine the packets, but they're from port 
443 and this is an SSL connection, so I can't really see what's going on.  Is 
there something in Mozilla which will show the HTTP headers that it gets?

oh.. i didn't realize this was an HTTPS link (doh!).. there are two options:

1) use about:cache to locate the cache entry for this page.. if you find the
   entry created with a "ClientID = HTTP-memory-only", you should be able to
   examine the HTTP headers there.

2) make a debug build and generate an HTTP log (see the file nsHttp.h for     
   instructions).
Thanks for the tips Darin, I found the page in the memory cache.  Here's
what it says:

key:
https://ais2.tiaa-cref.org/cgi-bin/WebObjects.exe/SeeAccum.woa/35/wo/7i30003O200Bw300Dl/0.2.1%20cache-control%3A%20private%20cache-control%3A%20max-age=0

fetch count:

2

last fetched:

Thu Jun 28 19:57:41 2001

last modified:

Thu Jun 28 19:54:40 2001

expires:

Thu Jun 28 15:53:10 2001

Data size:

23854

Security:

This is a secure document

Client:

HTTP-memory-only

request-method:

GET

response-head:

HTTP/1.1 200 OK
Server: Netscape-Enterprise/3.6 SP2
Date: Fri, 29 Jun 2001 02:52:12 GMT, Thu, 28 Jun 2001 22:53:12 GMT
Pragma: no-cache
Content-Type: text/html cache-control: private cache-control: no-cache
cache-control: max-age=0
Expires: Thu, 28 Jun 2001 22:53:12 GMT
Content-Length: 23854


  It looks to me like it's a bad Content-Type header, there should be a
semi-colon ";" after the text/html.  However, other browsers don't seem
to have a problem with it.  Mozilla 0.9 didn't.  [That "cache-control:max-age=0"
is on the same line as Content-Type but this text widget is making it wrap.]
More info:

On the site there are other links.  This is the headers from a link that
works correctly:

response-head:

HTTP/1.1 200 Apple
Server: Netscape-Enterprise/3.6 SP2
Date: Fri, 29 Jun 2001 03:09:44 GMT, Fri, 29 Jun 2001 03:10:41 GMT
Pragma: no-cache
Cache-Control: private, no-cache, max-age=0
Content-Type: text/html
Expires: Fri, 29 Jun 2001 03:10:41 GMT
Content-Length: 12453


I'm seeing this too on linux 0.9.2. What appears to be happening is this.
Mozilla is interpreting 

text/html cache-control: private cache-control:no-cache cache-control:max-age=0

as the MIME type, while obviously "text/html" is the only part of this that is
the mime type.

I don't know if this is any help. Also setting dependency on bug 78106 since
this is a MIME issue, at least at some level.
Blocks: 78106
Target Milestone: mozilla0.9.3 → mozilla0.9.4
Bad HTTP header.  this looks like it should be Evang...
it's either a bad HTTP header or mozilla is failing to parse a valid HTTP
header.  my guess is that is probably mozilla's fault ;-)

i suspect that the problem is that mozilla does not treat \r as an end-of-line
indicator.  instead mozilla accepts \r\n (per the HTTP/1.1 spec) and \n (since
many sites don't follow the spec).  seems we should also handle \r as well.

vance: if you could generate an HTTP log (see
http://lxr.mozilla.org/seamonkey/source/netwerk/protocol/http/src/nsHttp.h#32)
using a debug mozilla build, then i'd be able to confirm my suspicions.

thanks in advance!
Assignee: neeti → darin
-> mozilla 1.0
Target Milestone: mozilla0.9.4 → mozilla1.0
I'm not able to get a debug build on the OpenVMS platform.  If someone can get
me a debug build on VMS or Mac platform, I can try to get an http.log.  I don't
have the components to build Mozilla myself. Are there pre-built debug versions
on mozilla.org?
I don't know if this is because of a change on their end or on mozilla's, but
I'm not seeing this problem at this website with either 0.9.4 or 0.9.5

about:cache now says

key:
https://ais2.tiaa-cref.org/cgi-bin/WebObjects.exe/SeeAccum.woa/47/wo/pm4000oF600Z57009b/0.2.1
fetch count:

2

last fetched:

Tue 16 Oct 2001 05:48:51 PM CDT

last modified:

Tue 16 Oct 2001 05:41:43 PM CDT

expires:

Tue 16 Oct 2001 05:41:43 PM CDT

Data size:

24071

Security:

This is a secure document.

Client:

HTTP-memory-only

charset:

ISO-8859-1

request-method:

GET

response-head:

HTTP/1.1 200 Apple
Server: Netscape-Enterprise/3.6 SP2
Date: Tue, 16 Oct 2001 22:40:36 GMT
Content-Type: text/html
Content-Length: 24071
Status: NEW → RESOLVED
Closed: 23 years ago
Resolution: --- → WORKSFORME
marking WORKSFORME... please reopen if you can reproduce the problem.

also, you can now use a nightly build to generate a HTTP log.
Component: Networking → Networking: HTTP
Keywords: qawanted
QA Contact: benc → tever
Verified. 14 months without reopening.
Status: RESOLVED → VERIFIED
I'm not sure whether this was fixed in Mozilla or fixed on the TIAA-CREF
end.  Back then I sent them an email telling them that their HTTP headers 
were messed up.  The problem might have gone away if they fixed it (re: 
comment 14)
You need to log in before you can comment on or make changes to this bug.