>One of the key features of AFS which is really important to me
>as a site AFS administrator is its reliability.
I couldn't agree with you more Paul! And it needs to be to, as
dependency on AFS increase, need for controlled change management
increase... We find ourselves depending on AFS more each day.
>AFS software seems to have become more "solid and robust" over time.
Yes, like any other software that's been around that many years, and
have had fairly little fundamental changes applied to it lately. I do
not think however that this is the only reason, if you do not have
extreme control over the changes applied, even the tiniest change can
make the program go wild. We all know that. Judging from the stability
and the fact that changes _have_ been applied over time, Tranarc is
having, a good policy for applying canges to AFS. I find it hard to
belive that it's more than just good luck...
As I mentioned in my talk at Decorum: "Any change to a working
environment is a potential disaster, AFS is no exception to this"
We can only hope that the fact that AFS is back on track will not
make the developers turn wild and so excited they forget this simple
fact. Talking to some the AFS devguys at Decorum, I'm convinced they
will continue their good policy, and keep AFS stable through the process
of both maintenence and improvements of AFS.
>Encryption? There are many ways to do that already.
>Also, if encryption is your top issue, disconnect from the net
>and get into your Faraday cage.
Help! Help! Let me out! Let me out! ;-) The suggestion along this thread
is, not surprisingly, to add _another_ product from IBM; the firewall.
Without going into argument wether this specific product does the job,
I'm personally a little reluctant to solve a problem by adding another
product, generally speaking. Again, this could be a good suggestion, I'm
just a little conservative to this way to solve a problem:
New products adds complexity, and complexity means problems for any
sysadmin. During my years as a sysadmin this has become the challenge;
decreasing complexity. This is, beside the importance of AFS stability,
maybe the most important feature of AFS.
What I expressed my support for was the ability to easily switch on and
of someting that, by listening to some of the people along this thread,
appears to be in the system already. "Nothing is ever simple, this
simple fact often overlooked", [originated by my dear friend Lars Gunnar
Taube: http://www.init.se/~lgt]. This of course stands out true in this
case as well.
Encryption is not _the_ most wanted feature on the whishlist, and surely
there is more than one way to acheive this, but it is never the less on
my list. (Are you listening Transarc?)
>So (to Transarc), yes, do look to new capabilities (encryption,
>Episode, alternative authentication) but, whatever new stuff
>you bring in, please please don't degrade the excellence of AFS.
Again, I couldn't agree with you more on that one.
Thanks,
/peo
