Not being a true computer freak I have one, minor, quibble:
On Thu, 24 Feb 2005, Jeff Templon wrote:
> The idea is to make "name" the thing you really care about. I chose to go
> with Rod Walker's suggestion -- ELC for Enterprise Linux Compatible.
We have PCs with ~4GHz 64bit multicore CPUs with heatsinks that can hold a
burger.
Could we not just go _with_ "EnterpriseLinuxCompatible"?
[though I don't think it's as bad as "SL" - is that Scientific? or SuSE?]
> Let's face it: if we can't get a program to run the same on
> RHEL,SL,SLC,and CentOS, we are really screwed because these are supposed
> to be the same OS just recompiled by different people. [ Note I said
> "supposed" ... one could guess that if we can't get programs to run on all
> four, then someone broke the supposition ].
If it doesn't run on RHEL and x, then x is NOT "EnterpriseLinuxCompatible"
by definition and so should be a different OS name anyway, so we're not
that screwed. Except you do need an RHEL installation somewhere as a
reference!
On the subject of "how much should we reveal" I don't think we should be
too shy. Ahem. These descriptions apply to the Worker Nodes; my view is
that we should be going towards having the WNs totally sandboxed off from
the rest of the world with all interactions going through the local
service nodes... yet only monitoring (via R-GMA) is close - everything
else seems to be heading the other way. And I'm still worried that if the
wrong incident occurs, a knee-jerk reaction on-high might see the current
setup as a huge DDoS system and impose a solution we might not like.
Since it shouldn't be possible to hit a WN directly from outside, the
information system then wouldn't be revealing anything a malicious user
wouldn't get from 'uname -a'. The service nodes themselves could be
running something else completely...
Thanks
Henry
--
Dr. Henry Nebrensky [log in to unmask]
http://people.brunel.ac.uk/~eesrjjn
"The opossum is a very sophisticated animal.
It doesn't even get up until 5 or 6 p.m."
|