Subject: Re: /etc/mailer.conf vs. postfix vs. sendmail vs. ??? (was: CVS commit: basesrc/etc)
To: None <>
From: Greg A. Woods <>
List: netbsd-users
Date: 10/20/2002 14:31:47
[[ at best this dicussion belongs on netbsd-users ]]

[ On Sunday, October 20, 2002 at 12:49:44 (-0400), wrote: ]
> Subject: Re:  CVS commit: basesrc/etc
> And is there a "start here" document anywhere? Is there a "So, you
> just installed NetBSD. Awesome. Now fix email:" document anywhere?

Yes, there is:  Any generic guide to Unix systems administration.

Of course you need a little bit of common sense to go along with it, and
the ability to think about exploring and experimenting and _learning_
rather than just following cookie-cutter rules.

> Expecting my Mom or my brother to find the right man pages and read
> them is tantamount to expecting them to stick with MS-Windows.

If your Mom or your brother are not at least amateur systems
administrators then why would they even try to so set up e-mail
*services* on their own for themselves?!?!?!?!?

Just like Steve Bellovin's paper, and dozens or even hundres more
articles like it (indeed even a good half or more of Schneier's book
"Secrets & Lies"), say:  good systems administration is _HARD_.

Hiding what's really happening at the OS level behind a GUI (or worse
some delicate and often opaque thing like a "registry") doesn't make it
any easier.  Even Mac OS X, which makes some aspects of running a
desktop unix system quite trivial (at least to those of us who have a
deep understanding of what's really going on), doesn't provide all the
features for all the people -- and that's part of the big point here.

But then a desktop system has no business ever getting anywhere even
remotely close to the Public ("Big Bad") Internet.  Where's your Mom's
firewall?  Where are their network gateway servers?  If you answer any
questions like that with "on her one desktop computer" then you
instantly bring back the _desparate_ need for a highly skilled systems
administrator.  Now in theory the ISP should be providing at least the
option of a nice cosy and safe environment on which their cutomer's
desktop sytems can be used with access to supported Internet services.
Unfortunately with the ISP business being as cut-throat as it is the
majority of ISPs can only support the greatest common denominator and
that has turned out to be M$ (and maybe a glimmer of support for MacOS).
Obviously I'm blue-sky dreaming about even the relative safety and
service available to M$ users, but....

The point is that if you expect users with only average computer skills
to use a full general purpose operating system as a desktop system then
you really have to impress upon them that they have a great, and maybe
even desperate, need for "professional" assistance in doing the mundane
occasional systems administration tasks that such a system requires.

> Look, imagine the freshman at college who wants to run Unix on his
> computer.  A Linux install handles email just fine, FreeBSD and I guess
> OpenBSD would as well. Do we want that student to install NetBSD and have
> it be broken right out of the box? "NetBSD sucks! It can't even do email
> right!"

NetBSD isn't a desktop operating system, let alone a pre-configured one.

I don't know if generically that's a good thing, or a bad thing.  For me
it's a _really_ good thing in many ways.

If you, or your Mom, or your brother, or any other "average" user
without at least a modicum of systems administration skills wants to
have a desktop unix-based system that can safely and easily be used with
the average ISP then NetBSD is clearly _not_ the system of choice.
Certainly a suitably configured system with NetBSD at its core can be
created to fill this niche, but out-of-the box you cannot expect NetBSD
to be that system by default, not when it's trying to also meet the
needs of other very different classes of users.

> Imagine a business trying out different systems for use internally.
> Do we really want to create work for them just so they can use NetBSD
> similarly to how they would use any other Unix? What are they going to
> do? Realistically, broken email would be a strike against NetBSD. 

Realistically a business doing such an evaluation on its own without the
skills to understand what's going on is just plain _wrong_.  They are
only asking for trouble and will undoutably be unable to arrive at any
really meaningful result.

The fact that someone such as yourself would even assume that such a
thing might happen is a really sad indication of the mental state of
this industry and how people approach technology.

> A working system is better than a broken system with documentation.

Yes, but a properly working system only works in one specific situation
for which it was designed to work in.  Do you really want to create a a
system that only works out-of-the-box in some pre-concieved scenarios?
How many "big switch" flags can you put on sysinst before the complexity
overwhelms the "average" user?  Do you really want to create a system
that requires someone like an embedded systems developer, or an ISP, to
spend many extra hours undoing the glue that holds all the parts
together that are impossible to make use of in an embedded systems, or
ISP, environment?

This email issue is really a poor example on almost all fronts, but it's
the tip of the iceberg for the larger "integrated & packaged system
configuration issue.

After all, why the heck don't you just install Mozilla on your Mom's
machine and configure it for her in _exactly_ the same way as would have
to be done if she were using M$-Windoze?  Why would she ever need to run
a mail transport system on her "desktop"!?!?!?!?

I'm sure someone who would create a packaged and integrated desktop
system that used NetBSD at its core would be quite a hero amongst its
users.  I don't know if there's any incentive other than good will and
ego that'll make doing such a job worthwhile, but then again what do I
know about mass marketing?  :-)

After all this is _exactly_ the reason why there are so many GNU/Linux
"distributions".  Most try to serve the needs of a specific class of
users (and there are many in the "desktop" class).  Of course there are
some who try to wear so many different hats at the same time that all
they've done is to create something exactly in the image of that which
they are trying to battle against.  :-)  All these multi-hatted ones
have done is push the complexity out to another level and there's no
guarantee that the "average" user can deal with it any better while at
the same time there's ample evidence that it leaves those users with
unfilled, but very important systems administration requirements.

								Greg A. Woods

+1 416 218-0098;            <>;           <>
Planix, Inc. <>; VE3TCP; Secrets of the Weird <>