[comp.mail.elm] Elm Configure

baur@venice.SEDD.TRW.COM (Steven L. Baur) (06/28/90)

I have some questions and observations about the Elm Configure script.

The weirdest one is the combination of the questions about whether you will
run the resulting binaries on multiple machines and whether the hostname
should be compiled in.

Configure asks the latter question after you have already answered Yes to
the former.  This is probably a bug, but maybe unavoidable.

Who runs Elm on just one machine, and wants everything (including
~ pathnames) hardcoded?

Are there any installations out there that *have* to have the hostname
hardcoded?  The wording of that question (present in the Dave Taylor days)
sounds like he had an axe to grind somewhere.

Interesting is the fact that the domainname is hardcoded into the
configuration regardless, so I guess this precludes copied binaries of
elm into different e-mail domains.  This has some significance (to me)
because I cannot put out binaries for anonymous FTP within trw.com.

Does anybody care about this?

Just asking.

"If it ain't broke, don't fix it."
--
steve	baur@venice.SEDD.TRW.COM

taylor@limbo.Intuitive.Com (Dave Taylor) (07/03/90)

Steven L. Baur of TRW writes:

> Are there any installations out there that *have* to have the hostname
> hardcoded?  The wording of that question (present in the Dave Taylor days)
> sounds like he had an axe to grind somewhere.

Well, I'm confused.  I just looked over the Configure.sh script that
was in the last "Dave Taylor days" release (I like the sound of that! :-)
and I can't find anything to do with hardcoded hostnames.  

All that I can find that might be related is the question:

	Are you running a machine where you want to have a domain name 
	appended to the hostname on outbound mail [no] ?

and then a test to see if the system has 'gethostname()' available
(a call that *wasn't* available on all Unix boxes when I last had my
fingers in the distribution, at least.  It might be now, however)

Also, in the 2.1 release of Elm at least, you could simply tell the
program not to build the From: lines at all (DONT_ADD_FROM) and then 
it would let the lower level mail transport agent put in whatever was 
appropriate.  Which neatly avoids the problem of hardcoded domain 
names that you refer to...

I would be most interested in the 'axe grinding' prompts that you're
talking about, Steven!  Can you perhaps email me a copy of that bit
of the configure script?

			Straight from the stables,

						-- Dave Taylor
Intuitive Systems
Mountain View, California

taylor@limbo.intuitive.com    or   {uunet!}{decwrl,apple}!limbo!taylor

edwin@praxis.cs.ruu.nl (Edwin Kremer) (07/03/90)

In <642@venice.SEDD.TRW.COM> baur@venice.SEDD.TRW.COM (Steven L. Baur) writes:

   | I have some questions and observations about the Elm Configure script.

   | Who runs Elm on just one machine, and wants everything (including
   | ~ pathnames) hardcoded?
We do ;-)
Actually, we are of course running Elm on a lot of machines, but one
subset of that is a HP cluster with 31 diskless clients. In HP's
propriatary diskless protocol, all clients see *exactly* the same
filesystem as the server (unlike Sun's solution).
In this situation you can speak about "one system" and that's why
we've all the pathnames hardcoded.

   | Are there any installations out there that *have* to have the hostname
   | hardcoded?
Yes, there are ;-]
Same environment as above: none of the diskless clients are running a
sendmail daemon, so none of them will accept mail from another system.
The only system that does, is the HP cluster server. In this situation
it's very annoying to have the name of a diskless client machine appear
in the "From:" header line. What we want is that all diskless clients
pretend that they are the cluster server, so we really need hardcoding
the hostname of the server. (Yes, I know this can be fixed with
sendmail too!)


		hope this helps,
						--[ Edwin ]--
--
Edwin Kremer (SysAdm), Dept. of Computer Science, Utrecht University
Padualaan 14,   P.O. Box 80.089,  3508 TB  Utrecht,  The Netherlands
Telephone: +31-30-534104  | UUCP: ...!uunet!mcsun!hp4nl!ruuinf!edwin
Telefax  : +31-30-513791  | Email: edwin@cs.ruu.nl    [131.211.80.5]

peter@ficc.ferranti.com (Peter da Silva) (07/04/90)

In article <642@venice.SEDD.TRW.COM> baur@venice.sedd.trw.com (Steven L. Baur) writes:
> Are there any installations out there that *have* to have the hostname
> hardcoded?  The wording of that question (present in the Dave Taylor days)
> sounds like he had an axe to grind somewhere.

(raises hand)

We have a network of some 30 machines that look like one machine to mail.
I have had to make lots of changes to elm 2.2 to get it to work in this
environment. I don't plan on upgrading to 2.3 in the near term.
-- 
Peter da Silva.   `-_-'
+1 713 274 5180.
<peter@ficc.ferranti.com>

jwas@PacBell.COM (Joe Wasik) (07/04/90)

In article <642@venice.SEDD.TRW.COM> baur@venice.sedd.trw.com (Steven L. Baur) writes:
>"If it ain't broke, don't fix it."

"If it ain't perfect, improve it."

-- 
Joe Wasik - PacBell, 2600 Camino Ramon, 4e750, San Ramon, CA 94583  415-823-2422
jwas@PacBell.COM or {att,bellcore,sun,ames,pyramid}!pacbell!pbhyf!jwas
"Consitution, ... Shmoshtitution" -- Jesse Helms  (interpreted).

les@chinet.chi.il.us (Leslie Mikesell) (07/06/90)

In article <KZE4TMD@xds13.ferranti.com> peter@ficc.ferranti.com (Peter da Silva) writes:

>We have a network of some 30 machines that look like one machine to mail.
>I have had to make lots of changes to elm 2.2 to get it to work in this
>environment. I don't plan on upgrading to 2.3 in the near term.

Can you explain a bit more about how you did this and what problems
you encountered?  I'm considering something like this, perhaps using
smail 3.1 set up to lie in both the From_ and From: lines, and using
an alias file to resolve all user addreses.  As far as I can tell,
this would be transparent to the MUA (what did you have to do to ELM?).
Smail 3.1 can understand multiple names for the local host so there
should be no problem in accepting messages in local form or to the
common name and delivering to the specific machine from the alias list.
The main question I have is whether it would be better to modify
smail to be able to put different things on the From_ line, depending
on the destination, or to change all our external uucp connections to
use the MYNAME= feature of HDB uucp to actually look like a single machine
from the outside?

Les Mikesell
  les@chinet.chi.il.us

peter@ficc.ferranti.com (Peter da Silva) (07/06/90)

In article <KZE4TMD@xds13.ferranti.com> peter@ficc.ferranti.com (Peter da Silva) writes:
> We have a network of some 30 machines that look like one machine to mail.

In article <1990Jul5.175844.141@chinet.chi.il.us> les@chinet.chi.il.us (Leslie Mikesell) writes:
> Can you explain a bit more about how you did this and what problems
> you encountered?

The main problems hae been due to the fact that the C compiler we have is
broken. For the elm part of things, I compiled it with a fixed host name
and have it running through smail 2.5. I have smail 2.5 modified to insist
that all machines on the Xenix network are really "ficc.ferranti.com". I have
a GLOBALNAME define that is referenced in headers.c in 3 places: in generating
the default To and From lines, and where it checks for DOMAIN and hostdomain
to strip them off. Then I have a "users" file :include:ed from aliases that
maps the user name to a real user@system name.

I also hacked smail 2.5 to grok the % hack.

> The main question I have is whether it would be better to modify
> smail to be able to put different things on the From_ line, depending
> on the destination, or to change all our external uucp connections to
> use the MYNAME= feature of HDB uucp to actually look like a single machine
> from the outside?

All our external UUCP connections are through a single machine. I have smail
feeding a program (currently a shell script) in /usr/lib/smail/router that
figures what actual link to send mail through (network, internal UUCP, or
external UUCP).
-- 
Peter da Silva.   `-_-'
+1 713 274 5180.
<peter@ficc.ferranti.com>

baur@venice.SEDD.TRW.COM (Steven L. Baur) (07/07/90)

From article <935@limbo.Intuitive.Com>, by taylor@limbo.Intuitive.Com (Dave Taylor):
> Steven L. Baur of TRW writes:
>> Are there any installations out there that *have* to have the hostname
>> hardcoded?  The wording of that question (present in the Dave Taylor days)
>> sounds like he had an axe to grind somewhere.

> Well, I'm confused.  I just looked over the Configure.sh script that
> was in the last "Dave Taylor days" release (I like the sound of that! :-)
> and I can't find anything to do with hardcoded hostnames.

I stand corrected, sorry.  In e-mail to me, Syd attributed the wording of
this question to Larry Wall, and Brain Dead UNIX PCs as to the why of
having it there in the first place.  I misremembered when it first appeared.

I knew there was something wierd behind this Configure question, and had
forgotten that my first experiences with System V were with a version where
you had to relink the kernel to change the hostname (that system is sitting
on a shelf collecting dust, alas).


--
steve	baur@venice.SEDD.TRW.COM

6sigma2@polari.UUCP (Brian Matthews) (07/25/90)

In article <642@venice.SEDD.TRW.COM> baur@venice.sedd.trw.com (Steven L. Baur) writes:
|Who runs Elm on just one machine, and wants everything (including
|~ pathnames) hardcoded?

Me.  I've only got one machine, but even if I had a few, I'd trade some
additional (and bounded) upfront compile time for some (unbounded) time
later.

|Are there any installations out there that *have* to have the hostname
|hardcoded?

For some reason my SVR3/386 insists my hostname is 6sigma.sco.com.  I
haven't got around to figuring out why, so I hardcoded it to 6sigma for
now.  I guess that's a political or economic reason :-)
-- 
Brian L. Matthews	blm@6sceng.UUCP