[BusyBox] Busybox 1.0.0-pre10 and sendmail

[Rob Landley]

On Monday 10 May 2004 13:09, David Meggy wrote:
> On Mon, 2004-05-10 at 10:54, Michael Shearer wrote:
> > Hi,
> >
> > Why Sendmail, good question.  Well, it's the defacto SMTP daemon isn't
> > it?  I don't really need all it's features but apart from being a hassle
> > to configure, it's stable and secure.  I'd welcome any other
> > recommendations.
>
> I've found sendmail a bit of a pain, plus the fact it isn't really good
> for embedded server either.  If you check Debian, they have 5(?) mail
> servers.  I use exim for general use.  I just found it easier to use.

I'm looking for one myself.  Sendmail.cf has an entire o'reilley book devoted 
to it.  That's a bad thing.  This is not a book devoted to sendmail, this is 
a book devoted to its configuration file, which looks like more like line 
noise than obfuscated perl.  The majority of the code in sendmail deals with 
the intricacies of delivering mail via decnet, which doesn't even exist 
anymore.  Sendmail is the standard mailer the way ed is the standard editor.

I came to the conclusion it would be easier to write my own mailer in python 
than learn how to hack sendmail.cf.  (It can't be more than 1000 lines of 
code to handle incoming and outgoing smtp, plus bog standard file locking for 
mailboxes.  The biggest sources of complexity are 1) doing an mx record name 
lookup, and you can shell out to "dig" for that, 2) making sure you don't run 
out of disk space and that you've properly synced the data to disk before 
acknowledging receipt so you don't lose mail in case of a power failure.

Oooh.  Aaah.

Of course you could also use postfix.  Or exim.  Or qmail, except it was 
written by Dan Bernstein with all the problems that entails...

But pairing sendmail with busybox is officially peverse, yes. :)

> Mount should never change permissions!!  If those aren't the permissions
> you want, then you should change them on the NFS server.

Did the original mount command do this?  (I'm not saying we should copy them 
if they do something wrong, I'm just thinking we should start documenting 
known divergences at some point, even when we have a good reason for them...)

> David

Rob