Re: RE: AD request / L2 Triggers Chapter Statement

On Tue, Jul 09, 2002 at 05:20:22PM -0400, Phil Neumiller wrote:
> 
> Protocols for the Internet are not to be replaced but MAY surely be improved for
> wireless devices right?  

If it's not backwards compatible with the existing protocols, it
sounds like a replacement.  I think specific proposals are needed to
discuss this point further.

 
> If you don't agree; can you explain to me how QoS for wired/wireless links is going
> to work properly without L2 trigger events providing feedback to/through IP?  
> How about fast and seamless handovers for wireless devices to from other wired
> and wireless devices.

Well, I'm in the camp that needs to be convinced that fast mobility
isn't better handled at the link layer.  


> B).  The physics of wireless links is different from that of fixed links.  It is different 
>        enough to warrant at least some specialized treatment.

My understanding is that link technologies can overcome many of the
physical limitations of wireless networks (e.g., BER).  We've looked
at this in detail in draft-ietf-pilc-link-design-11.txt.  The place
where things seem to break down is where handoffs are occurring
between different link layer networks (e.g., 3g to 802.11).  However,
this isn't a limitation of the physics of wireless networks.  

On the other hand, if you are talking about the fact that your big ftp
stalls when you go into a tunnel with your wireless handheld.  Well,
do you really think this is big enough problem to rearchitect the
TCP/IP stack?  The Internet was design to support subnet heterogeneity
and specifically avoided optimizing for specific technologies.
(http://www.isi.edu/newarch/DOCUMENTS/rtb.IPAM.mar02.pdf has a
discussion of the design principles of the Internet that you might
find interesting.)  Basically, the "specialized treatment" you suggest
is best applied locally, i.e., at the link.


> E).  TCP is currently broken for wireless (I know there is work on
  going).


I find this statement surprising.  "Broken" is a very strong word.
Where is this documented?  We've been working on TCP performance over
wireless networks in PILC for two (three?) years now and this is the first time
I've heard it described as broken.  Perhaps your definition of broken
(& wireless, for that matter) is different from mine.  Anyways,
references would help.  


Regards, 


--aaron

_______________________________________________
pilc mailing list
[email protected]
https://www1.ietf.org/mailman/listinfo/pilc
http://www.ietf.org/html.charters/pilc-charter.html
http://pilc.grc.nasa.gov/