On 3/17/16 6:20 PM, Melinda Shore wrote: > On 3/17/16 2:38 AM, Stephen Farrell wrote: >> I always think of it as the end to end argument, not principle, >> and from that perspective, I think it remains entirely applicable. > > Yes, I do as well, but the IETF has not always responded > pragmatically to the ways networks are being deployed today > (and "cloud to cloud" isn't the issue). I started thinking it was an argument, and then John Wroclawski forcefully corrected me. My view these days is that it is an efficiency principle. It's not that nothing happens in the middle, but rather that things must go where they are most efficiently deployed from a systemic perspective. Where is it necessary to implement something in the middle, such as perhaps an IGP or a BGP, there's nothing wrong with that. Similarly, DDOS protection is best done as close to the source. If that's the network, so long as it doesn't cause other failures, great. NAT has survived because by and large it hasn't so seriously harmed end to end that the network couldn't be useful. Though it certainly has broken its fair share. I think pseudo-SMTP aware firewall may have done as much damage. I view the Tor case as interesting because they're a bit of a hybrid. There's a bit of application and a bit of network. The exit-node is semantically aware of JSON but strips it out. Interesting case study. Eliot > During the 90s, > ideological purity on the part of a number of participants > and at least one IESG member prevented us from responding > well to the NAT situation. > > But here's a problem: ideological purity and adherence to > good design principles tend to look like one another and I'm > not sure that it's always possible to tell one from another > except in hindsight. Another problem is that sometimes the > "right" way to solve a problem, at least within our framework, > doesn't work well with network operators' business models. > > I do think that one way to start to address some of this is > to reshape the way the organization is structured so being in > a leadership role (that is to say, the ones most likely > to be in position to block publication of a document they don't > like and to charter new work) isn't a full-time job, so that > people whose actual job it is to build networks, talk to > customers, and so on are able to step into those positions. I > don't think that will fix the problem but I think it would be > an incremental improvement. > > Melinda > >
Attachment:
signature.asc
Description: OpenPGP digital signature