56630 - Wrong character encodings set by Mozilla for this URL.

Status: VERIFIED WONTFIX

(Core :: Internationalization, defect, P3)

Description

[hirata masakazu]

To reproduce:Load the URL above.
Result:No background image, no frame, no table, but just some weird bunch of
characters are displayed.  The encoding is set to UTF-16 by Mozilla.  If you
change the encoding manually to Western or Japanese, the page gets redrawn
correctly.  However, hitting reload falls back to the initial state, even though
the character coding does not change back to UTF-16.
Reproducibility:Always.
2000101320Trunk and 2000101014M18
View Source or Page Info did not give me much help to sort out what is the cause
of this trouble.

Comment 1

[hirata masakazu]

The HTML at the URL has a metatag;
<meta http-equiv="Content-Type" content="text/html; charset=x-sjis">.
Opening the URL in Composer results in the same.  Interestingly, you can
manually switch the character codings to Western(ISO-8859-1) in the editor, and
then the file becomes editable.
See also bug 56626.
Switching component to Internationalization.

Comment 2

[Loco]

Fixing typo in summary.  used to say "charcter."  Also updating QA Contact and 
Assignee to reflect the change in Component.

Comment 3

[nhottanscp]

I was able to see the page without Auto charset detection.
But after I selected Japanese auto detection then I got the garbage characters.

Comment 4

[nhottanscp]

Change platform to all since I was able to reproduce with WinNT.
I think this problem is related to charset auto detection.
Reassign to shanjian, cc to ftang.

Comment 5

[Shanjian Li]

I spent some time on this problem. I found that the 1st byte of the message 
body is '\0'. When charset detector  try to detect its charset, the only 
possible explaination is le_utf16. That is before meta tag is even read. 
I am not sure if this problem is caused by our network code or it is the 
mistake of the website. 

Comment 6

[Frank Tang]

Do we know why the first byte is zero ? Can someone contact the webmaster?

Comment 7

[Shanjian Li]

The home page of this website (www.webstar-s.com) does not have this zero in text stream from
the http server. It does not seems like the http server's problem. I guess that index.html file
is started with a zero. I will close this bug. If anybody believe either:
1, This is not a website's problem, the extra zero was added by browser, OR
2, This is a common practice and many websites have this kind of behavior,

Then you can reopen the bug.
 

Comment 8

[hirata masakazu]

 I have emailed the webmaster about the URL above, but have not got any response.
 BTW, I have found another site that has one byte in its HTML that causes Auto 
detect to fall into UFT-16: 
 http://www.dontaku.com/table/fuzokuindex.html
 This one does not have the particular byte at the beginning, but in a tag in the 
Header of the HTML(yes, it has entire HTML in the Header, though).  The HTML does 
not have a meta tag to specify any charset, but I consider an error in HTML like 
this one should be taken care of by the Browser so as not to generate display of 
totally unrelated information.
  I am just submitting this information, and leave the status as WONTFIX.
If you guys think this is relevant, please reopen.
2000102008 MacTrunk.

Comment 9

[hirata masakazu]

One more point. View Source of the URL below is also affected by AutoDetect
switching to UTF-16.
Is this a correct behavior?  It seems strange that Browser and View Source
result in the same display.

 http://www.dontaku.com/table/fuzokuindex.html

Comment 10

[Teruko Kobayashi]

Verified as wontfix.