[rrg] NTP and first packet delivery

[David Conrad]

On Nov 30, 2008, at 12:45 PM, Tony Li wrote:
>> Recent postings suggest drop the first packet during map
>> lookup. Oeps!!!!
>> I think first packet delivery is a MUST.
>>
> The problem is that first packet delivery requires buffering.  To  
> quote an
> esteemed colleague:  "Buffering bad."  ;-)

Our esteemed colleague is over-generalizing (or perhaps is being taken  
out of context).  Buffering is not always bad, particularly since  
without it, the Internet probably wouldn't work.  In the context of  
performing a mapping lookup, it is part of a cost/benefit tradeoff.   
As has been said on numerous occasions, TANSTAAFL.  If you're going to  
have a mapping, you either need to take the hit of schlepping the  
entire map around to ever device that needs to do a lookup or you take  
the hit of increased latency as you buffer/drop the first packet and  
fetch the map entries you need.  There are numerous variations and/or  
gymnastics you can play to reduce the pain of the hit, but you'll  
_always_ have someone pointing out worst case edge scenarios where one  
solution is better than another.

You can argue that this indicates that adding a layer of indirection  
is fundamentally flawed, yet I have yet to see another approach  
proposed that passes the giggle test.

To be honest, this is all getting quite boring.  It has been over 2  
years since the AMS workshop and we're largely arguing about the same  
things that were raised in that meeting.  As far as I can tell, all  
we've been able to confirm is that all the potential solutions suck in  
some way and that they can't/won't be deployed operationally because  
of one particular sacred cow or another.

It would seem the future holds the following for us:
- increased PI allocations, leading to:
- increased cost in getting those PI allocations routed, leading to:
- increased use of NAT, leading to:
- increased use of overlay networks (e.g., "everything-over-HTTPS")

And of course, IPv6 merely exacerbates all of this.

I for one am coming to accept and welcome our not-so-new NAT  
overlords...

Regards,
-drc

P.S. Just for fun, I've appended the presentation I gave in AMS on 18  
Oct 2006.  Tell me if anything has changed.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: indirection.pdf
Type: application/pdf
Size: 122066 bytes
Desc: not available
Url : <http://www.irtf.org/pipermail/rrg/attachments/20081130/c71f2514/attachment-0001.pdf>