Skip Menu |
 

Download (untitled) / with headers
text/plain 7.9KiB
From hartmans@MIT.EDU Fri Sep 27 02:48:21 1996
Received: from dragons-lair.MIT.EDU (DRAGONS-LAIR.MIT.EDU [18.177.1.200]) by rt-11.MIT.EDU (8.7.5/8.7.3) with SMTP id CAA05673 for <bugs@RT-11.MIT.EDU>; Fri, 27 Sep 1996 02:48:20 -0400
Received: from MIT.EDU (SOUTH-STATION-ANNEX.MIT.EDU [18.72.1.2]) by dragons-lair.MIT.EDU (8.6.13/8.6.9) with SMTP id CAA20837 for <krb5-bugs@dragons-lair.mit.edu>; Fri, 27 Sep 1996 02:48:19 -0400
Received: from STARKILLER.MIT.EDU by MIT.EDU with SMTP
id AA25100; Fri, 27 Sep 96 02:48:19 EDT
Received: by starkiller.MIT.EDU (5.x/4.7) id AA29092; Fri, 27 Sep 1996 02:48:18 -0400
Message-Id: <9609270648.AA29092@starkiller.MIT.EDU>
Date: Fri, 27 Sep 1996 02:48:18 -0400
From: hartmans@MIT.EDU
Reply-To: hartmans@MIT.EDU
To: krb5-bugs@MIT.EDU
Subject: Consider upgrading to new Autoconf directory standards before 1.0 release
X-Send-Pr-Version: 3.99

Show quoted text
>Number: 37
>Category: krb5-libs
>Synopsis: Consider upgrading to new Autoconf directory standards before 1.0 release
>Confidential: no
>Severity: serious
>Priority: high
>Responsible: hartmans
>State: closed
>Class: doc-bug
>Submitter-Id: unknown
>Arrival-Date: Fri Sep e 02:49:01 EDT 1996
>Last-Modified: Thu Nov 14 22:07:19 EST 1996
>Originator: Sam Hartman
>Organization:
mit
Show quoted text
>Release: 1.0-development
>Environment:

System: SunOS starkiller 5.4 Generic_101945-37 sun4m sparc


Show quoted text
>Description:


Newer versions of Autoconf provide several different
directories with well-defined emeanings. For example, only user
executables are installed in sbin; daemons are installed in libexec.
Also, shared data like krb5.conf would be installed in share instead
of lib. (I propose to keep /etc/krb5.conf hard-coded as the first
place to look).

TO the extent that is reasonable, we should adopt these
conventions.


Show quoted text
>How-To-Repeat:

Show quoted text
>Fix:

Show quoted text
>Audit-Trail:

State-Changed-From-To: open-analyzed
State-Changed-By: hartmans
State-Changed-When: Fri Oct 18 01:14:12 1996
State-Changed-Why:
In general, this is as simple as changing pre.in and osconf.h.
We lose the functionality of seeting install_exec_prefix, etc, but
retain destdir so I don't think this is a major loss--we trade a
broken local hack for compatibility with the rest of the world.

Responsible-Changed-From-To: krb5-unassigned->hartmans
Responsible-Changed-By: hartmans
Responsible-Changed-When: Fri Oct 18 01:17:24 1996
Responsible-Changed-Why:
I'm implementing it.

State-Changed-From-To: analyzed-feedback
State-Changed-By: hartmans
State-Changed-When: Wed Oct 30 19:16:02 1996
State-Changed-Why:
Interested in discussion on changes that have been committed.



From: Sam Hartman <hartmans@MIT.EDU>
To: krbdev@MIT.EDU
Cc: krb5-bugs@MIT.EDU, ghudson@MIT.EDU
Subject: Re: krb5-libs/37: Consider upgrading to new Autoconf directory standards before 1.0 release
Date: Wed, 30 Oct 1996 19:15:53 -0500

I will soon be checking in some fairly major restructuring of
where krb5 files get installed to correspond more closely with more
recent file system standards and what Autoconf expects. Major changes
include:

* prefix/lib/krb5.conf ==> sysconfdir/krb5.conf (prefix/etc normally)
* prefix/lib/krb5kdc ==> localstatedir/krb5kdc (normally
* prefix/var/krb5kdc)

Marc had convinced me that these changes were a bad idea,
because while they tended to introduce compatability with other
systems and enabled several configure-time options, they broke
compatability with the Betas. He suggested I should use these options
only if there was a way that I could get the default behavior to be
what it used to be; no such way exists. However, when I was
complaining to Greg about this, he pointed out that making such
changes would only become difficult after 1.0 was released, and it is
really a bad idea to clutter up prefix/lib. I decided to make these
changes in parallel with having the debate about whether they are a
good idea, because they are fairly simple to revert, and I would like
a chance to work out the bugs if we do accept them.

It should be possible to add an
--enable-old-directory-structure option that sets things back the way
they used to be, but there are some problems that make it a fairly bad
idea for this to be the default. The main problem is that you would
then force the user to both disable this option and set whatever
overides to the directories they wanted if they wanted to change
something. I.E. configure lines might look like:
configure --disable-old-directory-structure
--localstatedir=/secure/krb5-database


instead of:

configure --localstate dir=/secure/krb5-database

This type of default behavior is more confusing than it's
worth.

--Sam


From: Sam Hartman <hartmans@MIT.EDU>
To: hartmans@MIT.EDU
Cc: krb5-bugs@MIT.EDU
Subject: Re: krb5-libs/37: Consider upgrading to new Autoconf directory standards before 1.0 release
Date: Thu, 31 Oct 1996 00:20:43 -0500

The following files were changed:
config/pre.in
include/krb5/Makefile.in
include/krb5/stock/osconf.h

From: "Theodore Y. Ts'o" <tytso@MIT.EDU>
To: Sam Hartman <hartmans@MIT.EDU>
Cc: krbdev@MIT.EDU, krb5-bugs@MIT.EDU, ghudson@MIT.EDU
Subject: Re: krb5-libs/37: Consider upgrading to new Autoconf directory standards before 1.0 release
Date: Thu, 31 Oct 1996 11:24:39 -0500

Date: Wed, 30 Oct 1996 19:15:53 -0500
From: Sam Hartman <hartmans@MIT.EDU>

Major changes include:

* prefix/lib/krb5.conf ==> sysconfdir/krb5.conf (prefix/etc normally)

I'm less concerned about this because I consider /etc/krb5.conf to be
the "standard" place most people should be using. The botch in Kerberos
V4 where some packages look for krb.conf in /etc/athena, some in
/usr/athena/etc, some look in /usr/local/lib, some in /usr/cygnus,
etc. was and is a mistake. We should have left it in /etc/krb.conf and
called it a day. Otherwise, you have all sorts of headaches when
different binaries look in different directories. Woe betide you if you
have more than one krb.conf file installed in these different
directories, and you wonder why one program works, and another
doesn't.....

So compatibility problems caused by moving krb5.conf from prefix/etc to
prefix/lib is just a specific case of a more general problem of putting
krb5.conf in lib in the first place.

I understand the rationale that is caused by people wanting to keep
krb5.conf available over the network (say, NFS, etc.). They can do that
by putting a symlink from /etc/krb5.conf to wherever they want krb5.conf
to live. I will note that the assumption that prefixdir is located over
NFS has its own entertaining security implications. :-)

* prefix/lib/krb5kdc ==> localstatedir/krb5kdc
(normally prefix/var/krb5kdc)

This is a KDC issue only, and so it's much easier to deal with.

Sam, I agree with you and Greg for checking this change in. I don't
believe we should both with an --enable-old-directory-structure.

- Ted

State-Changed-From-To: feedback-open
State-Changed-By: hartmans
State-Changed-When: Fri Nov 1 18:48:23 1996
State-Changed-Why:

Now that it's implemented and agreed to, it
should be documented.


From: Tom Yu <tlyu@MIT.EDU>
To: hartmans@MIT.EDU
Cc: krb5-bugs@MIT.EDU, krbdev@MIT.EDU
Subject: Re: krb5-libs/37: Consider upgrading to new Autoconf directory standards before 1.0 release
Date: Sun, 10 Nov 1996 21:58:22 -0500

While we're at it... /etc/v5srvtab is probably not such a great name
for the default keytab location. For starters, "v5" isn't all that
clear a nams, and "srvtab" is a holdover from krb4 days. How about
something like "/etc/krb5keytab" or "/etc/k5keytab"? Understadably,
if we're going to do this name change it should happen before 1.0.

---Tom

State-Changed-From-To: open-closed
State-Changed-By: hartmans
State-Changed-When: Thu Nov 14 22:06:34 1996
State-Changed-Why:

I believe this integration and the corresponding doc fix have been
successfully integrated.

Show quoted text
>Unformatted:
Currently, I do not let includedir set where includes are
installed, as our includes are architecture-dependent.