[1061] in Kerberos_V5_Development
Re: [Doug Engert ] Krlogind and ss-962301
daemon@ATHENA.MIT.EDU (Sam Hartman)
Sun Apr 14 17:14:39 1996
To: Marc Horowitz <marc@MIT.EDU>
Cc: Sam Hartman <hartmans@MIT.EDU>, krbdev@MIT.EDU
From: Sam Hartman <hartmans@MIT.EDU>
Date: 14 Apr 1996 17:14:34 -0400
In-Reply-To: Marc Horowitz's message of Sun, 14 Apr 1996 17:02:53 EDT
>>>>> "Marc" == Marc Horowitz <marc@MIT.EDU> writes:
Marc> Are you sure that *nothing* (no OS feature, no freeware, no
Marc> third-party commercial software, no anything) depends on the
Marc> way the vendor does things? If you start using streams, ps
Marc> output changes in the tty column. I am not ready to say
Marc> that this will not break anything. Are you?
I don't think we have ever claimed to have
interface-indistinguishability from vendor code. When you write an
Autoconf'd program, you are trying to abstract functionality tests
avoiding tests that depend on a particular system. In this style of
programming, you want to avoid system-dependencies when possible. As
a result, there may be minor differences in ps output and such things,
because autoconf will not always choose the same way of doing things
as vendor code when multiple options are available. "We're doing
things differently from the vendor," would be an acceptable reason to
change Imake templates, but alone, it is insufficient justification to
change configure tests. "We're doing things differently, and it
breaks in the following way," is often justification to make a change.
Marc> Marc
