[c-nsp] PA-POS-2OC3

[Robert E.Seastrom]

Rodney Dunn <rodunn at cisco.com> writes:

> And if at some point in the future you upgrade and
> the setup stops working because modules are powered
> down automatically don't be suprised.  
> That documentation is provided for a reason.

With all due respect, perhaps a rehash of third party GBIC debacle is
called for here.  Cisco didn't exactly win any friends in that one by
turning people's working configurations into non-working
configurations with a software "upgrade"...

> Just like when people try removing PA's from
> a GEIP and putting them in different VIP's.
> Those VIPs should power down. As an example:
>
> CSCed84692
> Externally found moderate defect: Resolved (R)
> VIP with unsupported PA should report PA unsupported and pwr it off

No, that's not the same thing at all, and I'm not asking you to make
my FDDI and 100-VG PAs work in a VXR either...

> The balance between too much flexibility and too much
> rope isn't as easy as it may appear at first glance.
>
> We made some mistakes when we even let the system come
> up in an unsupported configuration.

I know I'm not alone in running gigabit ethernets in colocation
facilities with substantially less than 100 MBPS on them.  Why?  Very
simple.  To reduce distance problems, I generally run fiber when going
outside the cage... and 1000-SX is more common than 100-FX.

I'll paraphrase Gert's warning about sending us to $VENDOR_J thus:

   If you insist upon protecting us from ourselves by enforcing
   backplane bandwidth quotas, then in order to remain logically
   consistent you must immediately discontinue sales and support of
   the PA-GE in combination with any NPE short of the G1.

12.2 and 12.3 print out one's bandwidth points when one types "show
hardware".  They complain upon bootup if they are loaded up too far.
That should be enough.

                                        ---Rob