mit-krb5 thread support -- fork safety

[Wyllys Ingersoll]

[ I'm taking this discussion to krbdev... ]

Sam Hartman wrote:

>>>>>>"Wyllys" == Wyllys Ingersoll <wyllys.ingersoll at sun.com> writes:
>>>>>>            
>>>>>>
>
>    Wyllys> OK, now I remember.  I'm going to take a look at your initializer
>    Wyllys> and allocator functions to see if we can use the same method.  I found lots
>    Wyllys> of places in the code where keys are manipulated in different ways,
>    Wyllys> so I didn't consider doing this originally.
>
>It's going to be a major headache.  There are lots of places as you
>point out where keyblocks are handled improperly.
>
>The hope is that the critical path key operations do not work this
>way.
>  
>

There are alot of places in the kerberos client utilities (kdb5_util, 
kadmin,
etc ) where keyblocks are used and manipulated which makes it even harder
to come up with a single interface for managing them. 

-Wyllys