Re: Questions regarding lsb-invalid-mta

[Aaron Sowry <aeneby@gmail.com>]

Hi,

> The lsb-invalid-mta was the closest we could get to a compromise. 
> Having the sendmail binary just disappear could cause scripts expecting 
> it to fail in unexpected ways.  OTOH, the sendmail binary has documented 
> error settings, and returning an error is a valid response.
> 
> It would seem that the package made its way into Debian in our effort to 
> harmonize the Debian and Ubuntu LSB packages.

Yeah, I had my suspicions that this was originally a downstream
initiative ;)

I disagree about the decision to use a "dummy" binary, though. Any
sensibly written software will handle a failed popen, but getting an
undefined error code back from something pretending to be the sendmail
binary is almost impossible to handle.

By the way, according to the sendmail man page, "Sendmail returns an
exit status describing what it did. The codes are defined in
<sysexits.h>". sysexits.h has the following define:

#define EX__MAX 78      /* maximum listed value */

So, it seems that 255 is in fact an invalid value to be getting back
from a call to sendmail. How is the caller supposed to interpret that?

> I agree with you that the proper thing to do here is to provide a proper 
> MTA when installing the "lsb" package.  Perhaps we should not build 
> "lsb-invalid-mta" in Debian as part of the "lsb" package, while perhaps 
> allowing it to be built on Ubuntu.

I definitely agree that this package should not be in Debian. If it's a
downstream problem, let downstream handle it. I have to admit that I'm
not entirely up to speed with Debian's packaging of it's MTAs however -
exim4 and sendmail appear to be mutually exclusive, and I see references
to "citmail" being a replacement for /usr/sbin/sendmail. Perhaps someone
with a bit more knowledge could suggest a possible solution?

/Aaron