> From: "Mike O'Dell" <mo(_at_)ccr(_dot_)org>
> the distinction is often based on the namespace used for making the
> forwarding decision
> the term "router" is *usually* applied to a device which is examining
> an L3 token ... a switch is usually examining a token other than an IP
> destination address - often an L2 address token
This is sort of the definition I prefer - although I tend to put it as "which
header are they looking at - the internetwork header, or the header used by
the access network".
(And, as Mo notes elsewhere, at one point there was a theory that boxes that
did "routing" were slower - a statement not entirely without reason, as a
router generally has to discard one local header, examine the internetwork
header, and create a new local header - certainly more work than just
examining a local header, and going. Unless you're allowed to throw some
modest number of mega-gates at the problem, yes, "routing" does tend to be
slower.)
Note that it may be hard to tell, from external examination, exactly which of
these two definitions (routing and switching) a given box is doing. However,
definitions which involve looking for mods to the packet may fail; I've known
"routers" which didn't decrement the TTL. And which TTL does one mean - MPLS
has a TTL, too, but an MPLS-only box isn't a router...
Noel