<p dir="ltr"><br>
On Sep 8, 2014 10:17 AM, "Adam Tauno Williams" <<a href="mailto:awilliam@whitemice.org">awilliam@whitemice.org</a>> wrote:</p>
<p dir="ltr">> I will agree that the replacement of ARP with discovery changes a lot of<br>
> things - way more than people realize [I suspect at this point most<br>
> people don't even think about it as they are so accustomed to the<br>
> idiosyncrasies of ARP].</p>
<p dir="ltr">His argument that ARP is well supported in hardware but multicast isn't is probably fair. Networking hardware has lagged in this area for years.</p>
<p dir="ltr">The network community has had an inappropriate dislike of multicast since it was first specified and many network admins and vendors avoid it to this day. One senior network admin told me that multicast was "Internet wide broadcast packets" and therefore would never be allowed on his network.</p>
<p dir="ltr">Campus scale IPV6 installs are fairly common in academia and I've never heard of one going sideways like this </p>
<p dir="ltr">And while it is convenient and sometimes unavoidable, extending all of your L2 domains to the core is asking for trouble even without IPV6. (Nothing more fun than watching a network loop take down the entire core switch.). The risk of network loops goes down substantially if you don't have a complicated spanning tree and if your north south links are L3. Don't think of spanning tree as a magnificent oak. It is more like Kudzu...<br></p>
<p dir="ltr">He doesn't go into details on the hardware but from his description i think they are using a layer 2 switch with rudimentary L3 abilities (probably a Ex8200)... This is acceptable for small networks but not a campus size one no matter what the sales guy says. Alas core routing hardware (juniper MX series is my fav) is fantastically expensive.<br></p>
<p dir="ltr">Mark</p>