Bug/Issue: References header doesn't wrap in emacs package

Bug/Issue: References header doesn't wrap in emacs package

[Allan Streib]

Notmuch/0.20.2 (http://notmuchmail.org) Emacs/24.4.1

Replying to a message in a deeply nested thread can trigger a complaint:

  sendmail: command failed: 550 5.7.1 Delivery not authorized, message refused: Message is not RFC 2822 compliant

The problem is a References header that is too long/not wrapped.

Regards,

Allan

Re: Bug/Issue: References header doesn't wrap in emacs package

[David Bremner]

Allan Streib <astreib@indiana.edu> writes:

> Notmuch/0.20.2 (http://notmuchmail.org) Emacs/24.4.1
>
> Replying to a message in a deeply nested thread can trigger a complaint:
>
>   sendmail: command failed: 550 5.7.1 Delivery not authorized, message refused: Message is not RFC 2822 compliant
>
> The problem is a References header that is too long/not wrapped.
>

Hi Allan;

Thanks for the report.  I can see how notmuch-reply would generate a
long references header in that situation. We rely on message-mode (part
of Gnus) to actually send the message, and it isn't clear to me yet if
message-mode (or some function it invokes) normally folds long headers.
So the first step is to figure out where to best fix this.  As a not so
nice workaround, if you remove References from message-hidden-headers
(using e.g. customize), you'll be able to hand wrap the references line.

d

Re: Bug/Issue: References header doesn't wrap in emacs package

[David Bremner]

David Bremner <david@tethera.net> writes:

> Allan Streib <astreib@indiana.edu> writes:
>
>> Notmuch/0.20.2 (http://notmuchmail.org) Emacs/24.4.1
>>
>> Replying to a message in a deeply nested thread can trigger a complaint:
>>
>>   sendmail: command failed: 550 5.7.1 Delivery not authorized, message refused: Message is not RFC 2822 compliant
>>
>> The problem is a References header that is too long/not wrapped.
>>
>
> Hi Allan;
>
> Thanks for the report.  I can see how notmuch-reply would generate a
> long references header in that situation. We rely on message-mode (part
> of Gnus) to actually send the message, and it isn't clear to me yet if
> message-mode (or some function it invokes) normally folds long headers.

I have reported this as an emacs bug.

  http://debbugs.gnu.org/cgi/bugreport.cgi?bug=21608

Re: Bug/Issue: References header doesn't wrap in emacs package

[Sanjoy Mahajan]

On 2015-10-02 21:21, David Bremner <david@tethera.net> wrote:

>>> The problem is a References header that is too long/not wrapped.
>>>
>>
>> Hi Allan;
>>
>> Thanks for the report.  I can see how notmuch-reply would generate a
>> long references header in that situation. We rely on message-mode (part
>> of Gnus) to actually send the message, and it isn't clear to me yet if
>> message-mode (or some function it invokes) normally folds long headers.
>
> I have reported this as an emacs bug.
>
>   http://debbugs.gnu.org/cgi/bugreport.cgi?bug=21608

I'm not sure whether fixing it in emacs is right.  The command 'notmuch
reply' is itself (with the sexp or json formats) generating the too-long
References: header.  Shouldn't it generate an RFC-compliant message?

Or should the json/sexp formats remain agnostic about line length,
because wrapping doesn't make sense with key/value pairs?  In that case,
I agree that message-mode should fix any long lines.

By the way, the latest Exim release (4.87), at least as configured in
Debian, does limit lines to 998 characters, and that limit has been
causing me a few hiccups -- usually when I reply to a message sent by
gmail with a zillion references (perhaps it lists every message in the
thread).

-Sanjoy

Re: Bug/Issue: References header doesn't wrap in emacs package

[David Bremner]

Sanjoy Mahajan <sanjoy@olin.edu> writes:


> I'm not sure whether fixing it in emacs is right.  The command 'notmuch
> reply' is itself (with the sexp or json formats) generating the too-long
> References: header.  Shouldn't it generate an RFC-compliant message?
>
> Or should the json/sexp formats remain agnostic about line length,
> because wrapping doesn't make sense with key/value pairs?  In that case,
> I agree that message-mode should fix any long lines.

For me the main issue is that the references header is editable by the
user before sending. So some input sanitization is needed at the sending
stage.

d

Re: Bug/Issue: References header doesn't wrap in emacs package

[Sanjoy Mahajan]

On 2016-06-07 06:28, David Bremner <david@tethera.net> wrote:

> Sanjoy Mahajan <sanjoy@olin.edu> writes:
>
>
>> I'm not sure whether fixing it in emacs is right.  The command 'notmuch
>> reply' is itself (with the sexp or json formats) generating the too-long
>> References: header.  Shouldn't it generate an RFC-compliant message?
>>
>> Or should the json/sexp formats remain agnostic about line length,
>> because wrapping doesn't make sense with key/value pairs?  In that case,
>> I agree that message-mode should fix any long lines.
>
> For me the main issue is that the references header is editable by the
> user before sending. So some input sanitization is needed at the sending
> stage.

Further supporting that argument, the user could also include a
1000-character line in the message body, which would require sanitizing
at the sending stage.

-Sanjoy

Re: Bug/Issue: References header doesn't wrap in emacs package

[David Bremner]

Sanjoy Mahajan <sanjoy@olin.edu> writes:

> On 2016-06-07 06:28, David Bremner <david@tethera.net> wrote:
>
>> Sanjoy Mahajan <sanjoy@olin.edu> writes:
>>
>>
>>> I'm not sure whether fixing it in emacs is right.  The command 'notmuch
>>> reply' is itself (with the sexp or json formats) generating the too-long
>>> References: header.  Shouldn't it generate an RFC-compliant message?
>>>
>>> Or should the json/sexp formats remain agnostic about line length,
>>> because wrapping doesn't make sense with key/value pairs?  In that case,
>>> I agree that message-mode should fix any long lines.
>>
>> For me the main issue is that the references header is editable by the
>> user before sending. So some input sanitization is needed at the sending
>> stage.
>
> Further supporting that argument, the user could also include a
> 1000-character line in the message body, which would require sanitizing
> at the sending stage.
>
> -Sanjoy

Headers are now folded in emacs master, apparently

        http://debbugs.gnu.org/cgi/bugreport.cgi?bug=21608

Re: Bug/Issue: References header doesn't wrap in emacs package

[David Bremner]

Allan Streib <astreib@indiana.edu> writes:

> Notmuch/0.20.2 (http://notmuchmail.org) Emacs/24.4.1
>
> Replying to a message in a deeply nested thread can trigger a complaint:
>
>   sendmail: command failed: 550 5.7.1 Delivery not authorized, message refused: Message is not RFC 2822 compliant
>
> The problem is a References header that is too long/not wrapped.
>
> Regards,
>

This bug should now be fixed in notmuch (commit 95e9c137e18a, will be
part of notmuch 0.24)

d