85999 - URL guessing overguesses

Status: RESOLVED DUPLICATE of bug 235786

(SeaMonkey :: Location Bar, defect)

Description

[Cesar Eduardo Barros]

From Bugzilla Helper:
User-Agent: Mozilla/5.0 (X11; U; Linux 2.4.5 i686; en-US; rv:0.9.1+) Gecko/20010612
BuildID:    2001061208

http://http:// (an obviously bogus URL) is somehow automangled to
http://www.http.com// 

"cross-platform, multi-language" (without the quotes; taken from the home page
and then pasted in content area) is somehow converted to
"http://www.cross-platform,%20multi-language.com/" (no quotes, again).

Can't the "add www. and .com" autoguessing misfeature be a little smarter, and
notice when its bogosity level is a bit too high? Spaces, commas, and a stray
colon should be a sutble hint of someone mistyping things, for instance, in a
href= attribute in a page, and shouldn't give bogus errors like:

"http://www.cross-platform, multi-language.com/ could not be found."

Reproducible: Always
Steps to Reproduce:
duh.

Comment 1

[timeless]

> Steps to Reproduce:
> duh.
please don't insult us.  the word middle is nowhere in your report and yet you 
expect a windows user to figure out that this requires middle clicking on the 
content area of a document.

imo the behavior of middle clicking is pretty clear, it means 'go to the url 
in the clipboard' if you disagree, and it appears you do, then we have a 
problem.  

http is a valid hostname, althought http: isn't a particularly valid anything.
ie5.5's first response is to hunt for host 'http' which i'm guessing is exactly 
what mozilla did.

of course, ie5.5 ctrl-enter for http://http:// 
leads to http://www.http://http://.com which amazing doesn't work.
i can imagine mistyping http://http// w/ the addition of a stray : (or the 
omission of say port 80).

anyways, after a long period of thought, i'm giving this bug to uid.

Comment 2

[Cesar Eduardo Barros]

The "duh" is because I need to put something in the "steps to reproduce" field,
or it won't open the popup (damn bug form quirks). And the steps to reproduce
are already clear: go to the URL in the "URL" field. Duh, it's obvious.

> the word middle is nowhere in your report and yet you expect a windows user to
figure out that this requires middle clicking on the content area of a document

No need to, just click in the "URL" link... 

The "taken from the home page and then pasted in the content area" bit just
explains how I did it; you can also type it in the URL field.

We shouldn't second-guess the user too much; first, http://<whatever> should
_never_ automangle to www.<whatever>.com (if you had the trouble to type the
prefix, you probably are already typing the full URL). Second, we shouldn't try
to correct the user; complaining that "this URL is invalid" is much better, and
makes him learn from his mistakes (instead of assuming "the computer will do
anything"). Third, we shouldn't _create_ bogus urls (which are what we are doing
in the second example).

Comment 3

[Boris Zbarsky [:bzbarsky]]

See also bug 85677 -- it's somewhat related.

Comment 4

[Keyser Sose]

Marking NEW.

Comment 5

[Blake Ross]

-> networking for resolution

Comment 6

[benc]

I suppose this is really a dupe of bug 16026.

Comment 7

[Michel Brabants]

When I enter "http://http//www.cebit.de/homepage_e" (this was a faulty link ...
on a page). It was sent to google to guess the page I think. However, I would
suggest that everything that starts with http:// is a link and it should'nt be
guessed. I can understand that this is a difficult toppic, but in this case it
wasn't the right page at all :)?

Comment 8

[Boris Springborn]

The following URLs are all diverted to http://www.microsoft.com/ :

http://http://bla.bla.bla
http://http
http

Mozilla/5.0 (X11; U; Linux i686; en-US; rv:1.7.5) Gecko/20041111 Firefox/1.0 
from debian sarge

Comment 9

[Lloyd Dewolf]

Bug # 235786 appears to be a dup of this bug.

Comment 10

[benc]

This is a dupe of 235786. I'm duping to a newer bug b/c that bug has traction,
and this bug has two problem descriptions.

The second problem (several words in the URL going to Domain Guessing, is
actually caused by bug 58867.

Comment 11

[benc]

*** This bug has been marked as a duplicate of 235786 ***