On Wednesday, 06/16/2010 at 02:02 EDT, David Boyes <[email protected]> 
wrote:
> I'd make a counterargument that if IBM intends to position z/VM only as 
a Linux 
> hosting environment, then the current setup that requires the user to 
have the 
> skills to rebuild CP to enable an ESM (which Chuckie usually recommends) 
is 
> highly user-hostile, and needs to be resolved at the CP layer, not in 
> third-party products. Every ESM shouldn't have to invent the CP 
interface 
> wheel. A cleaner, less intimately entwined, set of security interfaces 
would be 
> good for everyone, including the IBM products, and with the enormous 
amount of 
> work going into z/VM 6.next, this would be a good time to architect it 
in.

When building an ESM, you have some fairly complex challenges. At the top 
of the list, the ESM
1. Cannot be simply or easily circumvented (even by a sysadmin [as opposed 
to sysprog] ), 
2. Must enforce a limited set of operations when the ESM server is down 
sufficient only to get the ESM server back up and running or to declare 
"The ESM's dead, Jim."
3. Must recognize and differentiate system initialization activities from 
'steady state' stuff
4. Must generally support the idea the CP is the enforcement point. Policy 
may be derived from a server, but CP is where the Real Decisions are made 
and where it is understood that, sometimes, the policy does not apply or 
is a Really Bad Idea in some cases.

That's impossible to do if all the logic is in the server.

If the complaint is that ESMs are too hard to install and/or configure, 
then people who feel that way should open requirements.  Neither IBM nor 
CA need changes in the published CP-ESM interfaces to address such 
requirements.

Alan Altmark
z/VM Development
IBM Endicott

Reply via email to