<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 3.2//EN">
<HTML>
<HEAD>
<META HTTP-EQUIV="Content-Type" CONTENT="text/html; charset=iso-8859-1">
<META NAME="Generator" CONTENT="MS Exchange Server version 6.5.7652.24">
<TITLE>RE: [e2e] end of interest</TITLE>
</HEAD>
<BODY>
<!-- Converted from text/plain format -->

<P><FONT SIZE=2>Is there confusion, perhaps, between e2e as a philosophy and e2e as a necessity? I agree with Christian that a lot of the exciting new stuff happening is end-to-end. However: bit-torrent is e2e not because there are clean fundamental reasons for it to be so (though those may exist), but because that's the only deployment model open to the designer. I am not arguing that it shouldn't be e2e, just that nobody really had a choice to make it anything other than e2e.<BR>
<BR>
Now the follow-up question is --- as John Day pointed out, deployment difficulties have always been a fact of life, with the router/ISP companies being the primary reason back in the day. For me that raises the question: was e2e ever a clean philosophy, or was it simply the only practical deployment model open to designers twenty years back, as well? How many systems have been deployed end-to-end when people actually had a choice to do it some other way?<BR>
<BR>
And - do people have a tendency to use e2e the philosophy to incorrectly justify system designs that were e2e by necessity? (of course, sometimes the only deployment model turns out to be the correct one).<BR>
<BR>
(on the road currently, apologies in advance for late responses)<BR>
<BR>
- mahesh<BR>
<BR>
<BR>
-----Original Message-----<BR>
From: Rute Sofia [<A HREF="mailto:rsofia@inescporto.pt">mailto:rsofia@inescporto.pt</A>]<BR>
Sent: Mon 4/21/2008 5:34 AM<BR>
To: Christian Huitema<BR>
Cc: Mahesh Balakrishnan; end2end-interest@postel.org<BR>
Subject: Re: [e2e] end of interest<BR>
<BR>
Agree with Christian. And specifically focusing on L3, there seems to be<BR>
other wave coming, with FON and now the more recent Whisher. New<BR>
business roles (virtual ISPs, micro-ISPs) are also emerging.<BR>
<BR>
Rute<BR>
<BR>
<BR>
Christian Huitema wrote:<BR>
&gt; In what world are you living, exactly? What about systems like Skype, or<BR>
&gt; Bit torrent? They are definitely pushing the envelope of end to end<BR>
&gt; designs, are widely deployed, and are not controlled by major corporations.<BR>
&gt;<BR>
&gt;&nbsp;<BR>
&gt;<BR>
&gt; *From:* end2end-interest-bounces@postel.org<BR>
&gt; [<A HREF="mailto:end2end-interest-bounces@postel.org">mailto:end2end-interest-bounces@postel.org</A>] *On Behalf Of *Mahesh<BR>
&gt; Balakrishnan<BR>
&gt; *Sent:* Friday, April 18, 2008 5:11 PM<BR>
&gt; *To:* end2end-interest@postel.org<BR>
&gt; *Subject:* Re: [e2e] end of interest<BR>
&gt;<BR>
&gt;&nbsp;<BR>
&gt;<BR>
&gt; In fact there seems to be pushback from both ends --- we can't deploy<BR>
&gt; end-to-end protocols because major companies own the end-host stacks;<BR>
&gt; and we can't push mechanisms deep into the network because ISPs and<BR>
&gt; router companies own the network. Arguably the latter source of pushback<BR>
&gt; played a major role in the emergence of the e2e philosophy; but now we<BR>
&gt; have equally powerful commercial forces on the other side.<BR>
&gt;<BR>
&gt; So effectively the only practical mode of deployment seems to be the<BR>
&gt; 'almost' end-to-end middlebox --- one hop away from the end-host but not<BR>
&gt; quite into the network (and the Maelstrom work I presented day before<BR>
&gt; yesterday at NSDI would be one example).<BR>
&gt;<BR>
&gt; - mahesh<BR>
&gt;<BR>
&gt; --<BR>
&gt; <A HREF="http://www.cs.cornell.edu/~mahesh">http://www.cs.cornell.edu/~mahesh</A><BR>
&gt;<BR>
&gt;<BR>
&gt; -----Original Message-----<BR>
&gt; From: end2end-interest-bounces@postel.org on behalf of David P. Reed<BR>
&gt; Sent: Fri 4/18/2008 9:13 AM<BR>
&gt; To: Jon Crowcroft<BR>
&gt; Cc: 'end2end-interest@postel.org'<BR>
&gt; Subject: Re: [e2e] end of interest<BR>
&gt;<BR>
&gt; I personally think that the network community has become frustrated with<BR>
&gt; the inability to explore end-to-end protocols because the endpoint<BR>
&gt; stacks are &quot;locked in&quot; by vendors in proprietary code.<BR>
&gt;<BR>
&gt; &lt;snip&gt;<BR>
&gt;<BR>
<BR>
<BR>
--<BR>
Rute Sofia, PhD (rsofia@inescporto.pt, +351 22 2094263)<BR>
<A HREF="http://www.inescporto.pt/utm/~ian/">http://www.inescporto.pt/utm/~ian/</A><BR>
Area Leader/Responsa'vel de A'rea<BR>
IAN: Internet Architectures and Networking<BR>
UTM: Telecommunications and Networking Unit<BR>
INESC Porto<BR>
<BR>
</FONT>
</P>

</BODY>
</HTML>