[AFS3-std] rxgk-afs: moving SetCallBackKey to a separate document?

[Michael Meffie]

On Thu, 28 Feb 2013 14:57:00 -0500
Benjamin Kaduk <kaduk@MIT.EDU> wrote:

> Hi all,
> 
> From the previous discussions here, it really feels like section 10 
> ("Securing the Callback Channel") of this document is not nearly as 
> polished as the rest of the document.  That is, the current spec could not 
> possibly result in a working system.
> 
> Furthermore, the section is really describing a generic RPC and keying 
> system, with only a paragraph or two that are specific to rxgk operation. 
> We have several statements that begin "in rxgk's case, [...]" which makes 
> the section feel like a grafted-on afterthought.
> 
> rxgk's operation does not depend on SetCallBackKey, nor on extended 
> callbacks in general; on the contrary, extended callbacks depend on the 
> ability to set a callback key for a secure channel, and such a secure 
> callback channel is probably easier to impelement once rxgk is in place 
> than otherwise.
> 
> 
> Does anyone feel strongly that SetCallBackKey should remain in this 
> document?  I can probably take responsibility for writing the separate 
> document which would describe a SetCallBackKey RPC, if we do not just put 
> that behavior in the XCB document itself.


My impression is SetCallBackKey and the related topics could be part of a
Extended Callback document, because of the security considerations of extended
callbacks. If I understand, there is a not a case for securing callback
channels with the current callback RPCs since no information is leaked with the
current callback RPCs.

Best regards,
Mike


-- 
Michael Meffie <mmeffie@sinenomine.net>