[c-nsp] IS-IS question

[Saku Ytti]

On (2006-11-27 09:44 +0000), david.ponsdesserre at uk.bnpparibas.com wrote:
 
> - Different lifetime per database element:  IS-IS allows the association 
> of a specific refresh-timer per LSP (different from OSPF LSA).

However router might not accept TLV that has lifetime higher than
locally configured maxTLV of router.
Infact, if I read the standard correctly, this is how it (undesirably)
should work.
I know for a fact that JNPR and CSCO do not work like this, while
extreme does work like this. Only situation where this is relevant
is when migrating to higher lifetime, in extreme gear, you're
going to need flag day.

> - Flooding more optimal:  all IP prefixes are, with IS-IS, advertised 
> within one packet instead of generating one packet per database element 
> for OSPF.

Unless there are fragments.

> - Partial Route Calculation (PRC):  SPF (Dijkstra) is run when topology 
> has to be calculated (SPF tree) but when only IP routing information has 
> to be calculated IS-IS performs a Partial Route Calculation (PRC) which 
> consumes less CPU.

Although using iSPF in both might make this difference less important.

What I view as main advantage over OSPF
 - does not use IP (harder to attack)
 - no need to carry link IPs in IGP (lean and mean IGP is your best friend)
 - TLV's, very clean, very extensible
 - seems to get provider desired features first
 - probably better tested in large-scale provider backbones

What I view as main disadvantage over OSPF
 - does not use IP (harder to protect, if needed to)
 - availability in lower-end gear

-- 
  ++ytti