bug#26307: 25.1; smtpmail hard-coded auth mechanism order does not allow coping with misbehaving servers

bug#26307: 25.1; smtpmail hard-coded auth mechanism order does not allow coping with misbehaving servers

[Marco Maggi]

In GNU  Emacs 25.1.1 (x86_64-slackware-linux-gnu, GTK+  Version 2.24.30)
the library "smtpmail.el" contains the hard-coded list of authentication
mechanisms to try in sequence:

   (defconst smtpmail-auth-supported '(cram-md5 plain login)
     "List of supported SMTP AUTH mechanisms.
   The list is in preference order.")

it would be useful  to have a public API mechanism  to change such order
in case the server misbehaves when using the cram-m5 method.

  I am  customer of  a mail  service using a  server that  advertises to
support  the cram-md5,  plain and  login  methods but  then when  trying
cram-md5 the authentication fails; the login method succeeds.

  Just changing the above code to:

   (defconst smtpmail-auth-supported '(login cram-md5 plain)

and recompiling fixes my problems.

  It is hard  to contact the service administrators and  having them fix
the problem, so a workaround is useful.

TIA
-- 
Marco Maggi

bug#26307: 25.1; smtpmail hard-coded auth mechanism order does not allow coping with misbehaving servers

[Glenn Morris]

Marco Maggi wrote:

>    (defconst smtpmail-auth-supported '(cram-md5 plain login)
>      "List of supported SMTP AUTH mechanisms.
>    The list is in preference order.")
>
> it would be useful  to have a public API mechanism  to change such order
> in case the server misbehaves when using the cram-m5 method.
>
>   I am  customer of  a mail  service using a  server that  advertises to
> support  the cram-md5,  plain and  login  methods but  then when  trying
> cram-md5 the authentication fails; the login method succeeds.
>
>   Just changing the above code to:
>
>    (defconst smtpmail-auth-supported '(login cram-md5 plain)
>
> and recompiling fixes my problems.

Intead of recompiling, you can add to your .emacs:

(with-eval-after-load 'smtpmail
  (setq smtpmail-auth-supported '(login cram-md5 plain)))

Perhaps no Emacs change is needed then?

bug#26307: 25.1; smtpmail hard-coded auth mechanism order does not allow coping with misbehaving servers

[Marco Maggi]

Glenn Morris wrote:

> Marco Maggi wrote:

>>    (defconst smtpmail-auth-supported '(cram-md5 plain login)
>>      "List of supported SMTP AUTH mechanisms.
>>    The list is in preference order.")

>> it would be useful  to have a public API mechanism  to change such order
>> in case the server misbehaves when using the cram-m5 method.

> Intead of recompiling, you can add to your .emacs:

> (with-eval-after-load 'smtpmail
>   (setq smtpmail-auth-supported '(login cram-md5 plain)))

> Perhaps no Emacs change is needed then?

But the symbol is  defined as a constant, so it would  be bad style?  It
is  also  undocumented.  Ideally  one  should  be  able to  select  such
configuration on a per-server basis.
-- 
Marco Maggi

bug#26307: 25.1; smtpmail hard-coded auth mechanism order does not allow coping with misbehaving servers

[Glenn Morris]

Marco Maggi wrote:

>> (with-eval-after-load 'smtpmail
>>   (setq smtpmail-auth-supported '(login cram-md5 plain)))
>
>> Perhaps no Emacs change is needed then?
>
> But the symbol is  defined as a constant, so it would  be bad style?  It
> is  also  undocumented.  Ideally  one  should  be  able to  select  such
> configuration on a per-server basis.

It's (presumably) a constant because in normal operation there's no need
to change it, but if you have a special need to change it, it's easy to
do so.

Personally I don't see a need to add complexity to Emacs to cater to
misconfigured servers.