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

Re: [RRG] Idea for shooting down



I have well hi-lighted the advantages of my NIRA model before (no scalability problem, no churn,...)  In response to the current discussions, I should add: No changes in the hosts, no requirement that IPv4 addresses have to be  worldwide unique forever (i.e.IPv4 address depletion is no issue with this model).
 
By looking at Brian's concept: It is, by all its new details, just another form of hierarchy like Nimrod or PNNI, with a node aggregation which is a) cumbersome in doing and b) underperforming wrt its result.  In some older email I expressed my suspicion that the term "stretch" has been invented just to  discriminate hierarchy. Well, if you  only accept hierarchical models like Nimrod, PNNI, ISLAY, or Brian's new proposal than this criticism is appropriate.
 
Multihoming:
If I can route towards the destination (while ignoring any prefixes, but guided by geo.coordinates) as close as I want by design, than I can afford (!!) to treat sites (as defined by Brian) being nodes in my topology of lowest level s, whose adjacent links have such high weights that the site wouldn't be used for transit. (with the current BGP design no one can afford to extend the routing tables by let's say a factor 100 or 1000). But if that site is the destination node, IP forwarding may be done via any  of the site's adjacent links (to different ISPs) - i.e. multihoming is just multipath routing.
 
Postal mail versus email:
By looking at not so normal situation one may recognize whether a concept is sane or insane:
If I send a letter, let's say from Germany to Sausolito, but the receiver has moved to a different place, it is perfectly ok  that the last postal office would realize the receiver's relocation and would send the letter back.
It would be insane to disseminate that someone has changed his address to all postal offices in the world, just for the effect that an incorrect destination address could be detected by the ingress postal office.
I think the existing postal concept is sane, and of course, with the NIRA concept the email would also be sent to (at least California) and then returned. I think it would be ok if the email is returned with not deliverable. It would also be great if the first node who returns the email could also return the new address, if known,  respectively any other node on the way back  -  just as an assist of the DNS.
 
Imho, it could be so simple. Sometimes I compare my situation with Platon's cove story. Because I am not so good in BGP I am bad in interpreting the shadows on the wall, hence my model is ignored.
 
Heiner
 
 
In einer eMail vom 20.11.2007 06:11:51 Westeuropäische Normalzeit schreibt rw@firstpr.com.au:
With your proposal:

  http://www.cs.auckland.ac.nz/~brian/DFZng.pdf

I get the impression that the result would be the Internet being
much less meshed than it is at present.  This would have negative
impacts including:

1 - Less robustness in the event of link and router failure.

2 - Often longer path lengths (AKA "stretch") as a packet has to
    get out of one of the many Level 1 domains (or the one Level 0
    domain) up to some Level 2 domain in order to get to another
    Level 1 (or the Level 0) before it could be delivered.  This is
    true even if the ISP border routers connecting to the sites with
    the sending and receiving hosts are physically close, but happen
    to be of ISPs which are in separate domains.  (I assume an ISP
    can only be in one domain, but perhaps I am wrong.)

3 - Therefore, further dimensions of "Balkanisation" of the
    Net, where actual packet delivery times and reliability
    levels depend not just on physical (partly geographic)
    topology, link capacities and traffic levels but also on
    which domain each host is in and how far the packet has to
    travel to go up and then down a level to the destination
    domain.

4 - Therefore, a lot of fuss as ISPs try to decide which domain
    to be in.

5 - Also, perhaps, more pressure on "sites" (as I think you
    nicely describe them) to multihome to various ISPs in
    different domains.  However, that only helps with the
    path length if the "site" is smart enough to send outgoing
    packets to the optimal ISP's router.  This wouldn't help
    with optimising which ISP and region an incoming packet
    arrives through when it is sent by a non-multihomed site
    or a multihomed site without the smarts to choose the
    best outgoing router.


Also, I don't understand how your proposal would help with the
central problem we are trying to solve - of a multihomed site's
router requiring a separate route for every prefix advertised by any
site anywhere at all.

The fact that the destination site is in a different domain doesn't
alter the fact that the multihomed site's router needs a separate
route for that slice of address space, because the router still
needs to decide which of the two or more upstream ISP routers to
forward the packet to.

- Robin


--
to unsubscribe send a message to rrg-request@psg.com with the
word 'unsubscribe' in a single line as the message text body.
archive: <http://psg.com/lists/rrg/> & ftp://psg.com/pub/lists/rrg