Re: Questions regarding lsb-invalid-mta
Hi,
> I don't agree necessarily. I think that the issue is what kind of
> "compliance" you expect, binary compliance or merely "paper"
> compliance. Expecting a binary to be there that fulfills the sendmail
> ABI seems reasonable on the face of it. However, I can easily see a
> system being "LSB compliant" without a MTA, in fact, I think that
> might be a fairly common use case. So having a package that just tells
> the lsb checker to be less verbose would be fairly useful in my humble
> estimation.
A system is either compliant with a specification, or it's not. You
don't get to pick and choose. If the sendmail application installed by
lsb-core does not perform the functions required of it according to [1],
then the system is non-compliant. It's not partially compliant, it's not
"paper" compliant, it's just non-compliant.
That Debian tries to pretend that it *is* compliant despite this fact is
a bug IMO. The lesser of two evils must surely be to not provide
sendmail at all - that way, at least applications can figure out for
themselves what's wrong (sending mail is not always a user-initiated
function, and therefore error messages are not always seen by human
eyes). More from the package description; apparently sendmail:
"introduces a daemon which can cause security problems and second, users
get asked questions about how they want their MTA configured"
There's no requirement in the LSB that the system is able to receive
mail, just send it. Therefore, no daemon, no security problems, and no
configuration.
"The LSB requirement on /usr/sbin/sendmail comes from old times where
Linux and Unix machines had all fixed IPs and did server tasks in data
centers. Today's typical desktop Linux machines do not do local e-mail
any more as users use external e-mail services."
This part is obviously complete nonsense. I still see no reason why a
fully functional sendmail can't be provided - are there some limitations
with providing just the client in the way it's been packaged for Debian
or something?
> Whether the LSB is still relevant is another discussion and perhaps
> the more critical one.
I can agree with you here, and I'm happy to discuss this from the
viewpoint of someone developing software designed to run on LSB systems
(is there even an LSB reference system? is it possible to be compliant
at all)? But, for the moment, I think the more pressing issue is the
misleading behaviour of Debian (and derivatives) in this regard. It
needs to be addressed.
[1]
http://refspecs.linuxbase.org/LSB_3.1.1/LSB-Core-generic/LSB-Core-generic/baselib-sendmail-1.html
Thanks,
Aaron
Reply to: