Re: It's the Internet Stupid

[David Farber <dave () farber net>]

Begin forwarded message:

From: "David P. Reed" <dpreed () reed com>
Date: June 8, 2009 5:08:58 PM EDT
To: dave () farber net
Cc: ip <ip () v2 listbox com>
Subject: Re: [IP] Re:    It's the Internet Stupid

As a matter of technical accuracy, I respond to the following false  
statements from Brett Glass.

Brett Glass wrote:
> Since the information identifying (among other things) the protocol  
> being used is part of the packet contents, this would make it  
> impossible to prioritize time-sensitive traffic.
The protocol being used is not part of the packet contents.  The  
protocol is specified in the IP header, in an 8 bit field that  
specifies TCP, UDP, or ICMP, among other choices.  But in addition,  
the "time sensitivity" of IP packets is defined to be specified in the  
field of the IP datagram header called TOS.  This 8 bit field was  
chosen to specify this for a very important design reason, too.  Since  
new applications appear on the Internet every day, if not every  
minute, it would be an extremely *bad* choice to require routers and  
switches to understand every new application level protocols' needs -  
it would cause innovation to come to a complete HALT in the Internet.

(to be completely fair to Brett, the refusal by ISPs to agree to  
implement TOS, and to agree to support all protocol numbers (not just  
TCP, IP, and ICMP) has hampered innovation already.  But it seems like  
a weak argument that failure of ISPs to implement standards should be  
cause to pile non-standard, innovation-inhibiting kludges into the  
system. There is no "ruleset" that is acknowledged as correct in  
identifying which packets deserve priority treatment based on  
examination of "content").

The use of the word "impossible" should be a clue to the weakness of  
this claim. If prioritization can be achieved by implementing TOS,  
then the word impossible is wrong.
> Likewise, if providers could not route packets to a more direct  
> connection or send them at a higher speed when they were bound to or  
> from specific addresses, it would be impossible for a content or  
> service provider which required enhanced performance (e.g. low  
> latency or jitter) to pay a surcharge for higher quality of service.  
> This restriction (which would be the equivalent of prohibiting UPS  
> from offering "red," "blue," and "ground" service) would kill  
> innovation by precluding cutting edge technologies from ever seeing  
> the light of day. It would also effectively outlaw content delivery  
> networks.
The document Glass is commenting says nothing whatever about routing  
of packets to more direct connections or sending them at higher speed.  
All it says is that such decisions should not be based on who the  
sender is, who the receiver is, or what the content is.  Basing such  
decisions on what the sender *requests*, what the receiver *requests*,  
or what class of service the sender or receiver have *paid for* (or  
any other decision criterion not listed) would not be proscribed by  
words in the document.

Again the word "impossible" is just wrong.  A sender can make a  
request of more or less priority for a specific set of traffic in many  
ways, for example, both based on current standards (TOS, TIA1039, ...)  
or by standards yet to be invented (such standards are invented all  
the time - for example RSVP was invented a number of years ago, though  
it is not widely implemented).

I'd suggest that the discussion would be enhanced if discussants were  
to based their comments on technical facts, rather than strawman  
assertions of "impossibility".






-------------------------------------------
Archives: https://www.listbox.com/member/archive/247/=now
RSS Feed: https://www.listbox.com/member/archive/rss/247/
Powered by Listbox: http://www.listbox.com