[2055] in Kerberos_V5_Development
Re: comments on HP and BSDI problems
daemon@ATHENA.MIT.EDU (Sam Hartman)
Mon Dec 2 14:59:35 1996
To: "Theodore Y. Ts'o" <tytso@MIT.EDU>
Cc: Ezra Peisach <epeisach@MIT.EDU>, Sam Hartman <hartmans@MIT.EDU>,
krbdev@MIT.EDU
From: Sam Hartman <hartmans@MIT.EDU>
Date: 02 Dec 1996 14:58:44 -0500
In-Reply-To: "Theodore Y. Ts'o"'s message of Mon, 2 Dec 1996 13:46:29 -0500
>>>>> "Theodore" == "Theodore Y Ts'o" <tytso@MIT.EDU> writes:
Theodore> Here's a proposal. Tomorrow, we will make a single code
Theodore> thaw and refreeze, wherein we'll only put extremely low
Theodore> risk, small-impact changes for portability reasons.
Theodore> Things like #ifdef hpux are much safer than adding a
Theodore> configure test, unless we're willing to retest on all
Theodore> architectures, including those that we don't have direct
Theodore> access to. This will give us three days to rebuild
Theodore> binary distributions for a Friday 12/6 release.
Theodore> As long as we are putting in architecture dependent, low
Theodore> risk fixes, IMO this should be acceptable. What do
Theodore> other people think?
I'm mildly uncomfortable with this solution because it does
involve less testing of these patches than other changes that have
been made in the past. We will still have to recompile for all
architectures that we have built for, throwing away old binaries.
There for, you have most of the inconvenience of a full code thaw with
most of the problems associated with a quick fix.
Frankly, I don't see a problem releasing something that is
broken for Hpux along with a patch in the HP binary distribution.
Finally, by making patches instead of integrating these
directly into the release, we clearly indicate this is not a long-term
solution.
Theodore> - Ted
