77785 - URL: text2HTML: intl characters not recognized as end separators

Status: RESOLVED WONTFIX

(MailNews Core :: Backend, enhancement, P5)

Description

[hirata masakazu]

2001042508 Mtrunk
If a URL is in the Japanese parenthesis, clicking the URL sends the URL to
browser with three more bytes added at the end.
Thus, if there is a line in an iso-2022-JP message; 
?http://www.n-kan.com/kouen/2001/0407.html?
,the closing parenthesis gets underlined when viewed in Mail window. Clickng
this hyperlink results in a browser URL bar showing;
http://www.n-kan.com/seityo/20010420.html%EF%BC%89

This was not the case in the past builds.

Comment 1

[hirata masakazu]

Sorry for the mis-pasting. The original line in the example was:
?http://www.n-kan.com/seityo/20010420.html?

and the resultant URL in URL bar was:
http://www.n-kan.com/seityo/20010420.html%EF%BC%89

Comment 2

[nhottanscp]

I can reproduce this with NS6.01.
I tried a plain text mail.

Comment 3

[nhottanscp]

I think mozITXTToHTMLConv creats the link.
Reassign to [e-mail], cc to putterman, jenm.

Comment 4

[ji]

This problem still exists on 6.5 04/26 trunk build.

Comment 5

[ji]

This problem happens as long as there is a Ja character right after the URL, 
like http://home.netscape.comXXXX, here X stands for a Ja character.

Comment 6

[Ben Bucksch (:BenB)]

I suggest WONTFIX.

You cannot use any character that looks good to you to delimit the URL. The
"right" way is to enclose the URL in < and > (prefered) or double quotes
" and ". Whilespace is also acceptable, but more failure-prone. Everything else
that might work as delimiter is either
 - coincidence (because invalid in URIs)
 - heuristics adjusted to roman languages
   It's true that this is a bias, but we can't reasonably introduce heuristics
   for all languages of the world.

Unknown characters (not explicitly invalid in URIs and no heuristics) are
treated as valid characters of the URI. Thus, all caracters until a recognized
delimiter are considered part of the URI. This might include Japanese characters
directly following the Japanese quote (which, I assume, is != the roman ones).
These characters might be escaped later, which leads to the "%EF%BC%89".

Actually, it's a coincidence that this URI is recognized at all, considering
that there is a Japanese character at the beginning of the scheme. You could
argue that *this* is a bug.

Use < and > or at the very least whitespace around the URI.

References: RFC2396, especially appendix E.

Comment 7

[Ben Bucksch (:BenB)]

.

Comment 8

[hirata masakazu]

I do not get it, Ben. This is a plain text message. Nobody cares what RFC says
about correctness.

Comment 9

[Christopher Hoess (gone)]

The RFC in question is relevant because it defines the format and parsing of 
URLs.  Capsule summary of Ben's message: don't expect a URL to work properly if 
it's enclosed in Roman parentheses, Japanese parentheses, brackets, curly braces, 
or *anything* except angle brackets <> or maybe whitespace.

Comment 10

[Ben Bucksch (:BenB)]

Right. WONTFIX. Reopen, if you violently disagree. I'm also reassigning, in case
that happens (I don't plan to fix it).

Comment 11

[Ben Bucksch (:BenB)]

.

Comment 12

[benc]

-> mailnews/backend, per esther