Johannes Weiner <hannes@saeurebad.de> writes:

> On Wed, Sep 12, 2007 at 11:13:42AM +0200, Michaël Cadilhac wrote:

>> use something like this:
>
>> --- browse-url.el	12 Sep 2007 10:49:04 +0200	1.61
>> +++ browse-url.el	12 Sep 2007 11:09:27 +0200	
> [...]
>> -  (setq file (browse-url-encode-url file))
>> +  (setq file (browse-url-encode-url file "[*\"()',=;? ]" 'encode-percent))
> [...]
>> -  (setq url (browse-url-encode-url url))
>> +  (setq url (browse-url-encode-url url "[,)$]"))
> [...]
>> -  (setq url (browse-url-encode-url url))
>> +  (setq url (browse-url-encode-url url "[,)$]"))
> [...]
>> -  (setq url (browse-url-encode-url url))
>> +  (setq url (browse-url-encode-url url "[,)$]"))
> [...]
>> -  (setq url (browse-url-encode-url url))
>> +  (setq url (browse-url-encode-url url "[,)$]"))
> [...]
>> -  (setq url (browse-url-encode-url url))
>> +  (setq url (browse-url-encode-url url "[,)$]"))
> [...]
>> -  (setq url (browse-url-encode-url url))
>> +  (setq url (browse-url-encode-url url "[,)$]"))
>
> These use mostly the same argument.  Can't we generalize this?  Would it hurt
> the callsites if they all would use "[*\"()',=;? ]"?

Yes, it will.  A ``confusing char'' is just something Firefox or others
can consider as a URL separator or as a variable or something when the
website is passed as an argument to the executable, AFAIU.

`?=*' for example are not usually ``confusing''.  The only place those
chars are to be converted is when we browse for a file (thus when `?='
don't have their special meanings).

In the other cases, removing those chars destroys the meaning of the
URL.

-- 
 |   Michaël `Micha' Cadilhac       |  And please suggest to him that        |
 |   http://michael.cadilhac.name   |    he not refer to Microsoft Windows   |
 |   JID/MSN:                       |                  as "win".             |
 `----  michael.cadilhac@gmail.com  |          -- RMS                   -  --'