fragmenting the solution space

[Dave Crocker <dhc2@dcrocker.net>]

Folks,


As i recall being told, there was a great deal of debate about the way to add local area network support for TCP/IP.  After all, the characteristics of a LAN are far different from a WAN.  So, all sorts of optimizations are possible and appealing.

Instead of giving in to that urge, TCP and IP were kept uniform across the different environments.  The only concession to the LAN that showed up in IP, itself, was the ability to do direct exchanges, without requiring a router.

Oh.  That isn't really a LAN issue, is it?  It's an issue about direct-vs-mediated communication.  I guess someone looked for the abstraction that was relevant, rather than getting bogged down in the different details, creating divergent capabilities.  (And as usual ISO nicely provides examples of this mistake, in this case with TP[0-4].

It strikes me that mobility vs. multihoming is not benefiting from similar abstraction efforts, nevermind the likely artificial distinction between IPv4 and IPv6 on this topic and nevermind the real-world transition difficulties created by any suggestion that NATs much change.

Does it bother anyone that we are marching down a path that our next decade or longer will require support for at least:

1. IPv4 mobile
2. IPv6 mobile
3. IPv6 multihoming
4. TCP multihoming
5. SCTP multihoming
and so on?

I notice that my windows desktop has pretty awful performance with anything less than 512MB of memory.

It would be nice not to march the Internet stack down the path of requiring that much for my mobile phone.

  d/
  ---
  Dave Crocker
  Brandenburg InternetWorking
  +1.408.246.8253
  dcrocker  a t ...
  WE'VE MOVED to:  www.bbiw.net