Change in K4 ticket cache format

Jeffrey Hutzelman jhutz at
Tue May 17 22:30:04 EDT 2005

On Tuesday, May 17, 2005 06:08:17 PM -0700 Russ Allbery <rra at> 

> Jeffrey Hutzelman <jhutz at> writes:
>> I have a fiarly hideous proposal for a way to write out krb4 ticket
>> files on 64-bit platforms so that they will be parseable both by tools
>> that believe issue_date is 32 bits and those that believe it is 64 bits.
>> I also have corresponding code for reading it back in.
>> The downside is that my technique works only on little-endian platforms.
>> I'll post more details later; at the moment I'm off to dinner.
> I'm actually fairly convinced by the argument that it's not worth caring
> about, particularly after the discovery that we'd maintained local patches
> for years in this area.  So, at the least, don't worry too much about this
> on my account; I think we can deal with our problems locally for the
> remaining time that we're stuck with K4.  (We're getting some traction on
> getting the project to move off of it approved, so I expect that within
> three years we won't care any more.)

Well, when I say "proposal", of course I mean "running code I wrote months 
ago and have deployed".  See, we're starting to support Linux on x86_64, 
and wanted things we built against kth-krb4 to interoperate with the 
vendor's existing 32- and 64-bit krb4 libraries.

Take a look at /afs/

I believe it would be possible to write code which can read records written 
with a 64-bit issue_date even on big-endian systems.  But as I explain in 
the notes attached to the above patch, I've so far been unable to figure 
out how to write issue_date-length-independent files on big-endian systems. 
Fortunately, we don't care much.

-- Jeff

More information about the krbdev mailing list