[MlMt] Error?

[Bill Cole]

On 24 Mar 2014, at 16:23, Benny Kjær Nielsen wrote:

> On 24 Mar 2014, at 21:19, Benny Kjær Nielsen wrote:
>
>> I do have a Yahoo test account and I could try to find out exactly 
>> what makes the server return this error (using telnet/openssl),
>
> A quick test with MailMate shows that:
>
> * Sending a message with no subject and no content is not a problem.
> * Sending a message with a single URL as content results in the error 
> you are also experiencing.
>
> Based on the above, I still think it's a Yahoo issue and not a 
> MailMate issue.

Absolutely. However, it is a problem MM could handle better because 
site-unique reply codes are formally within the bounds of what SMTP 
servers can do and expect clients to handle properly. The flexibility 
has a long history:

http://tools.ietf.org/html/rfc821#page-48
http://tools.ietf.org/html/rfc2821#section-4.3.2
http://tools.ietf.org/html/rfc5321#section-4.3.2


Yahoo's SMTP server is a unique piece of software (originally a 
highly-modified derivative of qmail) with spam-detection hacks dating 
back to before there was widespread acceptance of the concept of 
content-based SMTP rejections or of any de facto standard model for how 
such rejections should be expressed. Due to the imbecilic mishandling of 
RFC1893/3463 enhanced status codes by MS Exchange, some MTAs with roots 
in the late 90s chose to define their own custom 5xy reply codes 
following the principles defined in RFC821 as a way of protecting the 
detail text from arbitrary replacement by the worst malware ever 
inflicted on Internet email. (Exchange...)

In practical terms, the right thing for a MUA to do when it gets a 5xy 
reply code at the end-of-data point that it does not recognize is to 
treat it as it would the generic 550 failure reply at that point or any 
other 5xy reply: return the message to an unqueued draft state and alert 
the user to its failure, providing the full unmodified server reply. It 
is reasonable to add the MUAs own interpretation of *STANDARD* reply 
codes and enhanced status codes, as long as there is clear distinction 
between "Your grumpy old mail server said THIS" and "Your friendly 
modern MUA thinks the problem is THUS".

FWIW, the handling of rejection during submission has been almost 
universally shoddy by all MUAs forever, so MM not being as clear & smart 
as it should be in this case is barely a critique at all. Eudora did 
reasonably well by v4...