[76672] in North American Network Operators' Group
Re: Anycast 101
daemon@ATHENA.MIT.EDU (Paul Vixie)
Mon Dec 20 11:45:08 2004
From: Paul Vixie <paul@vix.com>
To: nanog@merit.edu
In-Reply-To: Message from bmanning@vacation.karoshi.com
of "Mon, 20 Dec 2004 16:40:16 GMT."
<20041220164016.GD21109@vacation.karoshi.com.>
Date: Mon, 20 Dec 2004 16:44:40 +0000
Errors-To: owner-nanog-outgoing@merit.edu
> > Apparently you also didn't get any pointers to RFCs or other
> > authoritative sources that say "each and every packet injected into
> > the internet must be delivered in sequence".
>
> er... please quote chapter/verse here.
> these are "packets" and have sequence numbers
> -BECAUSE- they may not be received in order.
> the end-system must be designed to place the
> packets back in order before presenting the
> data to the application.
>
> e.g. this is not a circuit switched network.
of course it will work. it just won't be particularly fast. specifically,
it won't allow tcp to discover the actual end-to-end bandwidth*delay product,
and therefore tcp won't set its window size advantageously, and some or all
of the links along the path won't run at capacity. packet reordering is not
fatal to the technology, but it is fatal to the business. "not everything
that can be done, should be done."
