-----Original Message----- From: Khaled Omar Sent: Monday, October 2, 2017 11:43 AM To: 'Lee Howard' Subject: RE: [Int-area] FW: Request for a mailing list to IPmix I-D.
Lee, > You’re talking about updates to Microsoft Windows 10, 8, 7, and their sub > versions, Apple iOS (multiple versions), MacOS (multiple versions), Android, > ChromeOS, dozens of varieties of Unix (and *nix), plus all of the separate > lines of business supported by Cisco, Juniper, Alcatel-Lucent, Huawei, > Nortel. . . every network device manufacturer. It would take thousands of > work hours for each of those to get developed, documented, and tested. The number of technology companies and their OSs you listed is between 10 to 15 company, the software updates will take some time but of course not the same time as waiting enterprises globally to migrate to IPv6. > If even one of those vendors is unconvinced about your protocol, it will > fail, because your protocol requires every device on the Internet to support > it. True fact. > How do you define “a short time”? Because I still encounter routers (home gateways) without IPv6 support. A short time can be a year, and regarding routers/modems at homes will require firmware upgrade or replacement. > How will enterprise users’ enterprise networks get upgraded if the enterprise > operator doesn’t do it? They have to accept that updates to be able to connect to the Internet which is an easy task to do. Khaled -----Original Message----- From: Lee Howard [mailto:[email protected]] Sent: Sunday, October 1, 2017 1:37 PM To: Khaled Omar; Tom Herbert Cc: int-area Subject: Re: [Int-area] FW: Request for a mailing list to IPmix I-D. On 9/30/17, 12:54 PM, "Int-area on behalf of Khaled Omar" <[email protected] on behalf of [email protected]> wrote: >Hi Tom, > >I agree with most of what you have said, but changing the way of >encapsulation is not a hard work for running a code for all network >devices and hosts to be able to understand the new packet. That statement is insane. You’re talking about updates to Microsoft Windows 10, 8, 7, and their sub versions, Apple iOS (multiple versions), MacOS (multiple versions), Android, ChromeOS, dozens of varieties of Unix (and *nix), plus all of the separate lines of business supported by Cisco, Juniper, Alcatel-Lucent, Huawei, Nortel. . . every network device manufacturer. It would take thousands of work hours for each of those to get developed, documented, and tested. And they’ll be buggy. So network operators will carefully test before they roll it out. If even one of those vendors is unconvinced about your protocol, it will fail, because your protocol requires every device on the Internet to support it. > >There is difference between deploying IPv6 and implementing IPv6, IPv6 >was deployed in almost all OSs and network devices in a short time, but >implementing IPv6 by enterprises' users is the problem because there is >a dependence on them, they have to do it by themselves. How do you define “a short time”? Because I still encounter routers (home gateways) without IPv6 support. > >For IPmix, there is only deployment process that will be done by >technology companies, and enterprises' users will do nothing, they will >keep using their version of IP and regardless of the version they are >using, they will be able to access 100% of the Internet. How will enterprise users’ enterprise networks get upgraded if the enterprise operator doesn’t do it? > >Regarding running a code, there is no problem in that, a wg of people >will make things easier than letting only one do it by himself, >collaboration is requested. We’re both still looking for someone who wants to collaborate with you. Lee > >Khaled > > >-----Original Message----- >From: Int-area [mailto:[email protected]] On Behalf Of Tom >Herbert >Sent: Saturday, September 30, 2017 6:43 PM >To: Alexandre Petrescu >Cc: [email protected] >Subject: Re: [Int-area] FW: Request for a mailing list to IPmix I-D. > >> That said, I agree with you that that the technical basis is to allow >> encapsulate versions of IP in a same packet header. >> >This does not justify the statements that this protocol can be >developed and deployed in a short time by technology companies. >Deploying anything at Internet scale is hard, developing and deploying >a new IP version is at least an order of magnitude harder. Even if you >were to just change the version number of IPv4 and make no other >changes to the protocol, it still would be a stunning amount of work >that touches a myriad of devices and systems. End hosts need to change >several ways, NIC HW needs to change to support protocol specific >features, routers and switches need to be able to deal with new version >which most likely will require hardware changes, middleboxes and >firewalls need to deal with this, security needs to be fully >considered. Monitoring and diagnostic tools need to change, management >and administration need changes. But, on top of all that, there's no >magic switch that can turn up this new IP version on all hosts and >routers on the Internet instantaneously so we need to implement more >backoffs like happy eyeballs. Sorry, there is just no easy path here; >if there were it would have been applied to IPv6 long ago... > >> I would also rather say: the technical issue is how to use different >> versions of IP in a same IP packet. >> >>>> Since the initial posting of the IPv10, I and and others have asked >>>> several times for an implementation. >>> >>> >>> Yes, this have to be tested because theoretically it works fine >> >> >> I agree with you. Many concepts are first imagined in someone's >> mind, then sketched with a pencil on a paper, and discussed. These >> concepts work fine theoretically. >> >>> and I'm not a software developer to test it by myself. >> >> >> You are not. But can you make their job easy? >> >> Me too I can sketch a concept car on paper, but I doubt any >>manufacturer will ever make one :-) Because I dont know how to make >>cars. >> >IETF is "rough consensus and running code". It's incumbent on the >advocates of a proposed protocol to push for both of these. There's no >concept that protocol experts bring their protocols to IETF and then >throw over the wall to the protocol developers who happily go off and >implement the spec. If you don't have the skills for implementation, >then please find someone else to work with who can do it for you. > >Tom > >> Alex >> >> >> _______________________________________________ >> Int-area mailing list >> [email protected] >> https://www.ietf.org/mailman/listinfo/int-area > >_______________________________________________ >Int-area mailing list >[email protected] >https://www.ietf.org/mailman/listinfo/int-area > >_______________________________________________ >Int-area mailing list >[email protected] >https://www.ietf.org/mailman/listinfo/int-area > _______________________________________________ Int-area mailing list [email protected] https://www.ietf.org/mailman/listinfo/int-area
