Thus spake "Christopher Neill" <noise@cow.org>
> > "Portfast" and friends from other vendors will permit a loop to form
as
> > a transient and then clip it off.
>
> there cannot be a spanning tree loop on a port connected to one host.
You'd be amazed at how ingenious users can be.
User A connects a hub at his desk so he can use two PCs. So does user
B. Then A and B connect the hubs together "to make Quake go faster".
With portfast, there is a loop for a few seconds and then one port is
blocked. With no STP, your entire network grinds to a halt until you
manually track it down and pull one cable.
> like i said, none of my trunks, routers, ports connected to hubs are
> running portfast because i /do/ have redundant paths in my switching
> environment..
I'm sorry to hear that. You might want to fix that when you have some
spare time. (I'm not a fan of L2 switched environments)
> oh, yeah, heh, i don't let college kids near my switches..
I've seen more problems with sales types than college kids, actually.
> i'm assuming that this is the case for the person who posed the
original
> posit; obviously the safe thing to do with all ports is to not run
portfast
> on them until you know for sure what's going to be connected to them
> (and it helps to have some control over this)..
Control is an illusion; you can't stop users from doing stupid things.
I've found it's far more productive to *assume* that people will do
stupid things so I don't *need* to control anything.
> > Your intention that any particular port only goes to a "single host"
> > can be blunted by a visit to the Microwarehouse catalog. Any user
> > can show up with a repeater and go into the connectivity business.
>
> then why do i pay exodus several thousand dollars a month to give me a
> locked cage?
We're talking about end-users here, not colo centers. Look closely in
any office environment and you'll find mini-hubs stashed under desks and
behind cabinets all over the place.
S
This archive was generated by hypermail 2b29 : Sun Aug 04 2002 - 04:12:44 EDT