[lsb-discuss] Possible to get rid of /usr/bin/sendmail requirement in LSB 4.1?

[Jeff Licquia]

Martin Pitt wrote:
> Jeff Licquia [2009-06-12 13:31 -0400]:
>> Well, we have our deprecation policy.  In LSB 4.1, we can deprecate the
>> sendmail command, which would put it on track for removal around LSB 7.
> 
> You mean "sendmail being guaranteed to be available"? I don't think
> anybody proposed to deprecate the sendmail command itself, and I don't
> think it's a good idea.

The effect is the same, for two reasons: one, we're removing a
requirement, which we've promised not to do except within the
deprecation policy, and two, "optional" and "non-existent" in a standard
are effectively the same thing.

> But then the LSB fails to specify what happens if you call sendmail to
> send a mail which it cannot handle. In particular, if it just sends
> mail into oblivion, this wouldn't be a sensible interface for
> applications?

Actually, the LSB does specify that.  Per the specification, sendmail is
supposed to return nonzero when there's been an error.  That, at least,
tells the application that something's wrong.

Here is our spec for sendmail:

http://refspecs.linux-foundation.org/LSB_3.2.0/LSB-Core-generic/LSB-Core-generic/baselib-sendmail-1.html

> IMHO it is much better to have a semantics like "if /usr/bin/sendmail
> is available then you can use it" than the current "/usr/bin/sendmail
> is always available, but it's totally undefined what it does".

The "if /usr/bin/foo exists, you can use it" rule is implied in the LSB
already for any command not explicitly mentioned.