RE: [TERNLI] Notes from tonight's ad hoc

["Kevin Fall" <[e-mail]>]

Channel change might not be so exciting, but stuff like "I switched AP to a
different network" might be interesting.... and of course, data rate change
(e.g. 802.11b can support 4 rates).  802.16 has different convergence
sublayers it can use, which potentially affects mtu (gag).

Then there is the issue of L2 flow control, and the sort of curious stuff
linux does when it feels "locally" congested {where it basically acts as
though it recv'd a 'congestion experienced' indicator in an ack}.

- K

> -----Original Message-----
> From: Mark Allman [mailto:[e-mail]]
> Sent: Thursday, July 20, 2006 11:13 AM
> To: [e-mail]
> Cc: [e-mail]
> Subject: Re: [TERNLI] Notes from tonight's ad hoc
> 
> 
> > One thing that I took away from yesterday was that it may be
> > productive to build a pair of tables to describe the problem space;
> > one table for "significant path changes" that might be passed up to
> > the transport, and one table for "significant path desires" that the
> > transport can pass down.
> 
> This seems like a reasonable first step to me.  I would further winnow
> the list of things to be passed up as "characteristics that impact
> performance".  I.e., there is no use telling TCP that some random AP
> switched from channel 6 to channel 11.  TCP does not care.  Now, if that
> change results in more corruption or a lower bit rate then the TCP would
> care about the latter.
> 
> From there then we can organize it and figure out where the lowest
> hanging fruit really is.
> 
> My two cents.
> 
> allman
> 
>