From: Marco Stornelli <marco.stornelli@gmail.com>
Date: Tue, 04 Sep 2012 18:59:52 GMT
Subject: Re: [PATCH v4] Thunderbird: fix appp.sh format problems
Message-ID: <50464FA8.3020104@gmail.com>
In-Reply-To: <7vd321dc0e.fsf@alter.siamese.dyndns.org>

Il 04/09/2012 17:49, Junio C Hamano ha scritto:
> Marco Stornelli <marco.stornelli@gmail.com> writes:
>
>> 2012/9/4 Junio C Hamano <gitster@pobox.com>:
>>
>>> I would expect, at least when you are responding to an existing
>>> message, some of them are filled already (and if so, I think appp.sh
>>> wants to know exactly how, for example, has RFC2047 quoting already
>>> applied, or are we supposed to write in UTF-8 and let Thunderbird
>>> massage the contents when we give the file back to it?), and also
>>> there would appear In-Reply-To: field already filled (possibly there
>>> may be References: as well).
>>
>> Message reply is out of scope of my patch. The goal here is send a
>> patch, so the execution flow is to open a new message,
>> clik on external editor (configured properly), select patch file and
>> send. It was the scope of the old script and it is the scope of my
>> patch.
>
> I certainly can understand that you updated the script for that use
> case and that use case only, but given that the original tries very
> hard to preserve:
>
>   - what was in $HEADERS (by only replacing Subject);
>   - the recipients CC'ed in $HEADERS (by grabbing them into $CCS); and
>   - the body of the message in $BODY (i.e. what came after $SEP),
>
> I find it hard to believe that it was meant to work on a freshly
> created empty message and nothing else.  If people were depending on
> the recipients listed on Cc that are taken from $1 to be preserved,
> your patch will introduce a regression for them, no?
>

I think all the process is different. Current script rely on the user to 
fill Cc: and To: in message composition window, it does a union of what 
found in commit message as signed-off-by (adding own address in cc?!) 
and Cc (usually not filled in the commit message and we should even 
count acked-by, tested-by and so on). My vision of things: the user can 
create a patch *already* with all information using --to and --cc. If 
you want to add optional addresses, you can of course add them in the 
composition window. In addition, with this approach is really easy to 
use external script as get_maintainer.pl of linux kernel, load the patch 
and send, really easy. So I don't think it's a regression, it's only a 
change in the work flow.

Marco