[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: Proposed Resolution of Issues [1-37]

Fred Baker wrote: 
On Sep 2, 2005, at 1:33 AM, Tim Chown wrote:

Issues 32&33: ULAs *may* help renumbering, not *will*. There is baggage
with ULAs and as such their use is a tradeoff not a given, I feel.


I very much agree. I had the same debate with Pekka and Thomas when the renumbering procedure draft went in. Thomas filed a 'discuss' requiring me to say that ULAs would simplify renumbering, but without saying how. I told them:

On Jan 15, 2005, at 5:46 PM, Fred Baker wrote:

Listen. If ULAs simplify the procedure of renumbering a network without a flag day, then there should be several places in the document where a few sentences of the following form can be added.

"if the old prefix..." or "if the new prefix..."
"is a ULA prefix then"
"...this step may be skipped" or "...this step may be simplified <in this way>"
"and it still allows you to renumber a network without a flag day for <this> reason."

Present me with those sentences, and I will include them. 

I don't think any of that applies. My take is that ULAs have the advantage
during a renumbering exercise of providing business continuity for
*internal* operations - your printers, for example, can have ULAs
and are untouched (and their DNS entries are untouched) during the
renumbering process. Your internal SMTP system can keep
operating without any change to internal DNS. Etc. So it does take
a number of hosts off the renumbering list - exactly those
hosts that have no external visibility.

<snip>

If you can't tell me *how* ULAs reduce the complexity in the procedure of renumbering a network, I don't see the mention as relevant.


I have heard a fair bit of dogma about ULAs to the effect that they solve something, but have never been presented with those sentences, and am not convinced on the rest either. They look to me like a re- invention of site-local with many of the same problems. 

No. They don't have the problem of ambiguity. In context of massive
hosting centers for hundreds of customers, they are vastly better
than RFC 1918 addresses (which is the comparison my operational
colleagues make).

But you do have to accept that a large number of enterprise
networks *want* private address space precisely because it isn't
supposed to be globally routeable. That may not be intellectually
satisfactory as a reason, but it seems to be reality.

And by the way, if ULAs are global, they are global addresses... What does the "L" stand for? 

They're not global. They're globally unique but intended for local use.
However, "local" can mean a whole enterprise intranet plus VPNs to its
business partners. In that context, global uniqueness is a desirable
property.

    Brian