[1072] in Kerberos_V5_Development
Re: [Doug Engert ] Krlogind and ss-962301
daemon@ATHENA.MIT.EDU (Marc Horowitz)
Mon Apr 15 18:51:21 1996
To: "Richard Basch" <basch@lehman.com>
Cc: Sam Hartman <hartmans@MIT.EDU>, krbdev@MIT.EDU
In-Reply-To: Your message of "Mon, 15 Apr 1996 18:38:10 EDT."
<199604152238.SAA00639@badger.lehman.com>
Date: Mon, 15 Apr 1996 18:51:56 EDT
From: Marc Horowitz <marc@MIT.EDU>
In message <199604152238.SAA00639@badger.lehman.com>, "Richard Basch" <basch@lehman.com> writes:
>> If you mean to minimize changes from the vendor's shipped release,
>> well the problem is that the entire Kerberos distribution is not based
>> on the vendor's release....
I have had customers come to me and say "the case in this constant
string isn't just like the vendor's, this is breaking scripts." They
could care less if the entire code base was different (which, of
course, it was). They wanted it to act as close to the same as
possible. While Lehman may now be enlighted due to your input (good
for you and for them), most of the rest of the world is not
(unfortunately), and the kerberos team should recognize this.
>> I may be viewing it as a purist to try to not have the OS specific
>> tests, but I have seen the code work on a wide variety of platforms
>> without the OS specific tests, and if this can be continued, then it
>> helps support OS FooBar X.Y which might be released in 6 months.
Putting a line in configure.in to force use of bsd ptys instead of
streams for hpux 10.0 is not going to affect any future release of any
operating system. This argument hold no water, and you have repeated
it twice. I, too, have seen the configure tests work on many
platforms, but I have seen them fail on a few. The configure tests
are very good, but they are not perfect, or we would not be having
this discussion. If someone wants to sit down and write an AC_TRY_RUN
test case for streams vs bsd, instead of depending on the existence of
symbols like we do now, that would be even more robust and I would
support it, but nobody seems to be doing this. So, I suggest a
platform check as a cheap workaround. If we discover that many
vendors have broken streams, then perhaps the new test would be called
for.
Marc
