Ian MacLure([EMAIL PROTECTED])@Sun, Nov 27, 2005 at 06:31:28PM -0800:
> ------------------------------
> Message: 7
> Date: Sat, 26 Nov 2005 15:05:48 -0800
> From: Wade Curry <[EMAIL PROTECTED]>
> 
> <SNIPPET>
> MVS is stretching me again as I learn a new security system (RACF).
> I have to humbly admit that I am doing better with this than others
> who know the *system* better because I understand the *importance*
> of it, while some still are only annoyed &/or intimidated by it.
> (It's very much based on ACL's, by the way)
> <SNIPPET>
> 
>       RACF?
> 
>       Is that still around?
>       It was state of the art in the heavy iron world 20+ years ago.
>       Wasn't very secure IMHO. Apparently wasn't too difficult to
>       compromise according to some colleagues at the time.
> 
>       IBM

Yep, it's still around.  Security is very modular, or so I'm told.
RACF is the IBM product, though, so I think it is actually fairly
ubiquitous.

I can't compare it to older versions.  I don't know of
vulnerabilities other than the user-introduced variety.  At this
point I have to dance around the edges of my knowledge.  MVS
training these days is mostly on-the-job, learning from people with
strong experience and ability, but who have forgotten a lot of what
they knew. I'm not working with the SYSPROGs, so that is part of
it.  My knowledge has some good sized holes in it as a result.

RACF has only presented one issue that really bothered me.  The
issue is that I wanted to give a job access to only the one dataset
that it needed.  I defined that dataset's profile as
HLQ.WHATEVER.JCLLIB and gave the job read-only access.  Well, other
jobs and users needed access to that dataset, too.  However, they
were given their access using HLQ.** as the dataset profile.  The
effect is that only the most fully qualified profile will be used
to determine who gets access.  So my defining that profile actually
removed access that was granted in the more general one.

I understand the purpose.  It does make it difficult to
create a profile that doesn't wreck existing permissions.  It has
encouraged people to use the most general profiles they can, and
subverts the goal of having security in place.

Wade Curry
syntaxman


-- 
[email protected]
http://www.kernel-panic.org/cgi-bin/mailman/listinfo/kplug-list

Reply via email to