[7461] in Kerberos
Re: object file size
daemon@ATHENA.MIT.EDU (Sam Hartman)
Tue Jun 11 18:01:45 1996
To: schwartz@galapagos.cse.psu.edu (Scott Schwartz)
Cc: kerberos@MIT.EDU
From: Sam Hartman <hartmans@MIT.EDU>
Date: 11 Jun 1996 17:49:44 -0400
In-Reply-To: schwartz@galapagos.cse.psu.edu's message of 11 Jun 1996 15:59:55 -0400
>>>>> "Scott" == Scott Schwartz <schwartz@galapagos.cse.psu.edu> writes:
Scott> At the risk of harping on an issue of perhaps secondary
Scott> importance, please file this under performace bugs:
Scott> Binaries are quite a bit larger in beta 6 than in beta 5.
Scott> On SunOS, statically linked rlogin is more than 400K.
Scott> Telnet is more than 600K. Why so big? Shared libraries
Scott> only hide the problem; the bloat still exists, but in some
Scott> other file.
The problem is that Kerberos V5 comes with a
significant amount of functionality. In particular, the two binaries
you mention include both Kerberos V5 and Kerberos V4 libraries, as
well as code to deal with ptys, utmp, and user interactions.
A lot of interoperability issues were dealt with in Beta6.
The result is that servers and clients have to deal with a wide range
of potential options and responses from remote counter parts. In
addition, much more is user configurable than in previous releases.
There are areas where code size could be sinsignificantly
reduced, and many areas where performance could be increased
significantly However, if the current and expected future
functionality is to be maintained, don't expect the code size to
decrease much.
If you are looking for a Kerberos implementation for imbeded
systems or space-constrained environments, the MIT implementation is
not your best bet.
If you have specific suggestions on functionality that can be
removed or reorganization of code that would allow us to reduce size,
please let us know.
