Re: gnus / message-send-mail-with-mailclient [patch]

[David Reitter]

On 24 Mar 2006, at 14:29, Reiner Steib wrote:

> On Wed, Mar 22 2006, David Reitter wrote:
>
> [ Apparently you message didn't make it to the list (yet) => Not
>   trimming the quotes.  Please keep emacs-devel and address@hidden in Cc. ]

Strange thing, it's missing indeed. I did cc it to emacs-devel, but  
not to address@hidden  Oh well.

> ELISP> (setq sendmail-program "/foo/bar/baz")
> "/foo/bar/baz"
> ELISP> (setq message-send-mail-function 'message-send-mail-with- 
> sendmail)
> message-send-mail-with-sendmail
>
> ==> Searching for program: no such file or directory, /foo/bar/baz
>
> > Maybe we should caution users appropriately if mailclient is used on
> > systems where we can't guarantee reliability?
>
> I can't think of a reliable way to check if the application called by
> `browse-url' handled the mailto link correctly.

I can't either. I believe `message-send-mail-with-sendmail' is the  
best default for `message-send-mail-function', unless running on  
systems where sendmail is not implemented or swallows mail. (OS X  
10.4 is supposed to automatically fire up sendmail, according to an  
Apple developer. I haven't seen this work reliably, so far.)

> At least, `mailclient-send-it' could inform the user what it does.
>
> Something like "Delegating [shortened-mailto-stuff] to
> `browse-url'...".  Maybe including the value of
> `browse-url-browser-function' in the message.  Note that
> `browse-url-browser-function' might be a (REGEXP . FUNCTION) list.

These require developer-level knowledge and add to the general  
information overload. "Browse" never sounds like mailclient.
How about

(message "Delegated message to default mail client.")

But most importantly, we should see if the "Message sent" message can  
be avoided, since it's plain and simply wrong from the user's  
perspective.