2011/8/18, Nejc Škoberne <[email protected]>:
> Dear Gang,
>
>> That is not practicable. You can't do such modification on innumerable
>> hosts.
>
> Who says only "innumerable" hosts will want to have this? My view on this
> is,
> that these proposals are not meant to solve only a specific
> problem for a specific scenario. Why would we want that? If we do proposals
> this
> way, we will lost ourselves among very similar but a bit different
> standards.
> It's more complex to "administer" such set of standards and it is much
> harder
> for the developers to crawl through all these documents in order to
> implement
> what they need.

"NOT PRACTICABLE" has two meanings.
1) Restricted port aware host is relatively few case
2) Socket need to be modified. That is a practical challenge, not only
bind() but also connect()

>> Sorry. I still can't see much benefits from that.
>
> I'll try harder to deliver possible real-world scenarios:
>
> 1. Home Torrent node.
> ---------------------
>
> I am a customer of an ISP who deployed one of the A+P mechanisms. I have a
> home
> server which is also able to run Torrent client. And since this is a new
> server,
> it happens to support the A+P mechanism of the ISP. So I can plug it
> directly
> to my ISP-provided link (modem, CPE, switch, whatever), and everything /just
> works/. This way:
>
> - I don't have to deal with static RFC 1918 addressing,
>
> - I don't have to deal with static NAT configuration on the CPE,
>
> - I don't have to deal with with additional port configuration on my Torrent
> box.
>
> - ...
>
> It just works.
>
> 2. Enterprise server mode.
> --------------------------
>
> I am a network/system administrator for my company, which has a very limited
> set of public IPv4 addresses. But as I don't want to maintain a separate
> IPv4
> addressing infrastructure in my network, my server network will be IPv6-only
> as well. However, the servers still need IPv4 connectivity, and for that I'd
> like to use A+P. This way:
>
> - I don't need any special ALGs for my servers, which I would have to
> develop
> first since I am using my own proprietary protocols, which can't traverse
> NATs
> well,
>
> - I don't need to maintain IPv4 in my network where I don't want to have it,
> especially I don't need to maintain _another_ IPv4 addressing (RFC 1918),
>
> - I get rid of stateful NAT44 in the core, which is very difficult to make
> highly available properly (state synchronization).
>
> - ...
>
> ----
>
> Can you see the benefits now?

The consumption is to produce a brand-new network. If that is the
case, I may prefer going to native IPv6 network.


Gang
_______________________________________________
Softwires mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/softwires

Reply via email to