[1074] in Kerberos_V5_Development
Changes necessary for packet-mode streams
daemon@ATHENA.MIT.EDU (Sam Hartman)
Mon Apr 15 18:59:53 1996
To: "Richard Basch" <basch@lehman.com>
Cc: Marc Horowitz <marc@MIT.EDU>, Sam Hartman <hartmans@MIT.EDU>,
krbdev@MIT.EDU
From: Sam Hartman <hartmans@MIT.EDU>
Date: 15 Apr 1996 18:59:40 -0400
In-Reply-To: "Richard Basch"'s message of Mon, 15 Apr 1996 18:38:10 -0400
>>>>> ""Richard" == "Richard Basch" <basch@lehman.com> writes:
"Richard> On Mon, 15-April-1996, "Marc Horowitz" wrote to "Richard
"Richard> Basch, Sam Hartman, krbdev@MIT.EDU" saying:
>> In message <199604152203.SAA00603@badger.lehman.com>, "Richard
>> Basch" <basch@lehman.com> writes:
>>
>> >> The problem is I don't know how well HP-UX 10.0 works.
>> (While we do >> have some such machines, I don't readily have
>> access to them to test it >> out to comment further.)
>>
>> I had some serious problems when I tried to get streams working
>> on hpux 10.x, which is why I reverted to berkeley-style stuff.
>> Doug/Sam seems to have gotten streams to work, but I'm still a
>> little uncomfortable. I'm probably also biased from my OV
>> days, when many of our customers required the absolute minimum
>> changes from the vendor-provided programs. It would be
>> unfortunate for Cygnus or OV or any other kerberos vendor to
>> have to change the MIT release because we were trying to be
>> intellectual purists. This is nice for an educational
>> environment, but is inadequate for the real world. If we want
>> to make kerberos widely-accepted, it needs to be as painless as
>> possible, and this means minimizing unnecessary changes.
"Richard> With Sam's comment about the STREAMS mode not properly
"Richard> handling packet mode, we may already have a foregone
"Richard> conclusion, which is to use the BSD style interface in
"Richard> certain cases. As a result, if I have some time I might
"Richard> look into what it would take to support packet mode in
"Richard> the STREAMS implementation.
Telnetd already handles streams style packets if you want to
look at that as a basis. It would be nice if you could create
something like pty_read that dealt with both Berkely and streams style
packet mode, in some standard mode that deals with the possible return
values of both. Then, packet mode could be enabled with
pty_enable_packet_mode or something, which would try pushing pckt if
you have streams, else using the ioctl. I'm not sure how you would
deal with passing whether Berkely or streams packet mode ended up
getting used, but that could be dealt with.
I'll think more about this interface and send mail. I don't
expect that I'll have much time to implement before summer, and I
won't be around over the summer.
--Sam
