<<< Date Index >>>     <<< Thread Index >>>

Re: stuff_all_quoted



On Wed, Aug 16, 2006 at 07:16:13PM -0700, Gary Johnson wrote:

> No.  This is the sort of problem I was afraid we were going to get
> into; it's a case where what you think the message contains is not
> what RFC 2646 says it contains.  (I'm going to indent the following
> by four spaces to set it off and maybe avoid corruption.)  The line
> 
>     > > Blah blah blah blah stuff here
> 
> in a format=flowed message is considered quoted to a depth of one by
> the leading '>' and to have been space-stuffed with a single space.
> (See sections 4.2 and 4.5 of the RFC.)  The text is then

I totally understand that bit. But in this example, we're not talking 
about responding to a message that's encoded according to the 
format=flowed RFCs.... see below. 

> > And how can mutt really reliably "know" when text from a non-flowed 
> > message is quoted and when it's not.

> It can't.  Format=flowed quoting is well defined; everything else is
> just convention.

Right. And I think this is the heart of the problem.

Currently, mutt's "support" for flowed text extends exactly as far as 
saying "I know I'm generating flowed text, so set the content-type 
appropriately". I think mutt should either go further towards supporting 
generating messages with flowed text, or remove support for it entirely 
(I don't use the feature at all, so I don't really care one way or 
another myself). I think it's a really tricky task to accomplish without
integration between the MUA and editor.

Here's the problem.... when responding to a NON-flowed message with any 
previous quoted text, for the response to be correctly quoted, the 
(apparently) quoted text needs to be reformatted - otherwise 
everything's a big old mess. Every MUA I can think of that generates 
flowed text does this.

The mutt approach of "that's the editor's problem" doesn't really work 
because the editor has no way of knowing how the original message was 
encoded - it just gets a bunch of text dumped into it. I guess you could 
do some weird stuff in vim (or any other editor that might support this) 
to reformat accordingly, but that's a huge pain, especially since you 
wouldn't want to setup anything automatic since (again) the editor has 
no idea what type of message is being responded to.

I think the thing to do (ugly though it might sound) would be to (when 
$text_flowed is set) reformat the message in flowed style based on 
$quote_regexp (and let the user deal with any outside cases where this 
screws up the quoting).

> > But unfortunately, if mutt wants to try to implement format=flowed 
> > support, I think it makes sense for the program to try and do the 
> > best job it can at "guessing" whether parts of a non-flowed message 
> > are quoted when quoting it.

> I don't think mutt should be guessing.  You are free to edit a
> message in your editor any way you like.  If you want to have your
> editor guess at quoting levels and reformat replies per those
> guesses, that's fine too.  But I don't think mutt should be
> reformatting anything except as allowed by Internet standards.

That's ridiculous, though. The current behavior *doesn't* really enforce 
internet standards at all, because it makes it very easy for the user to 
break things by setting $text_flowed and then not generating flowed 
text, or not generating it properly. And leaving it up to the editor 
isn't really "fair" when the editor doesn't have access to the same 
information (most importatntly, how the original message was encoded) 
that mutt does.

AFAIK, there's no standard which says you *can't* reformat a message; 
the very act of quoting a message involves reformatting.

I'm pretty sure that almost all mailers which fully support generating 
flowed text (including Eudora, which is made by Qualcomm, who came up 
with the whole thing in the first place) attempt to quote text from 
non-flowed mailers (hrm - gmail doesn't do a good job of it!).  
Otherwise, we'd be getting messages with really mixed up quoting all the 
time...

I did some tests with Pine, and it does a pretty good job (well assuming 
whoever you're responding to isn't using a stupid quote string, anyway).  
The quoted non-flowed text is quoted consistently, but doesn't have an 
extra space at the end of the line as does the flowed text.

so:

  On such and such a date, Foo wrote:$
  $
  >> Blah blah blah blahblah. This text wasn't flowed when it was$
  >> quoted, but now it's quoted neatly. This is the message before$
  >> the one I responded to.$
  > $
  > This is the message I responded to $
  $
  And here is my text. Neato.$ 

Again, though, my biggest concern is that the 99.9% of mutt users who 
*don't* send messages using flowed text can view and reply-to flowed 
messages properly / neatly / consistently. Granted, all of these have 
their own editors as well, which puts mutt in a weird position. To be 
honest, I think it might be better to just support displaying it but not 
generating it.

w