Mechanism extensions and the GSSAPI

[Nicolas Williams]

On Mon, May 03, 2004 at 04:07:12PM -0400, Sam Hartman wrote:
> >>>>> "Nicolas" == Nicolas Williams <Nicolas.Williams at sun.com> writes:
> 
>     Nicolas> RFC2744 speaks to this issue.  While it doesn't specify
>     Nicolas> an ABI it does encourage implementors to specify one.
> 
> 
> Yes.  Within an implementation you will have an ABI.
> 
> I know of no platform that has standardized an ABI across GSSAPI
> implementations on that platform.

A case can be made (IIRC Martin made it) that the vendor of an OS
effectively sets the GSS-API C ABI for that platform if they ship a
GSS-API implementation.  So, effectively, Solaris as a GSS ABI, and so
does RedHat Linux, and...

More so if they have a pluggable, open mechglue (ours isn't open, but it
could be, and then I bet any interested 3rd parties would want to know
about the ABI without which the SPI wouldn't be open, and if the API is
the SPI...).

Nico
--