BG> As you can see, I connected at 26400, and although the entire call
BG> took only 11 seconds, the modems had transparently retrained upwards
BG> to an asymmetrical 28800 bps in both directions.  Interesting, wot?

RS> I think its distinctly arguable if its a very satisfactory way to actually
RS> do the line speed stuff. I find it extremely hard to believe that the line
RS> really does change, and mostly improve, in such a short time. They are in
RS> fact deliberately choosing to start to low for some reason.

BG> How do you explain the vast majority of calls which actually
BG> start out at the maximum link rate of either 28800 or 33600 then?

I'm not convinced that the 'vast majority' actually do.

BG> In fact, on those calls which have started low, I used to do
BG> an ATY11 line frequency response dump, and it was quite obvious
BG> that the high end was marginal for full speed connects.

Yes, but thats a separate question to whether you would expect
the line conditions to generally IMPROVE during the call. You
would normally expect that it was just as like to move down as
up with a session which is marginal for full speed connects.

I agree that clearly it should be using a lower rate with the session
isnt adequate for full speed, but increasing being more common is odd.

BG> Why it generally improves almost immediately is beyond me though.

RS> Even sillier with a V32bis call, the modems negotiate a 14400
RS> connect fine, and then the damned USR virtually IMMEDIATELY demands
RS> a retrain, on every single call I have made so far. Nuts IMO.

BG> Have you called many (or any) other USR V.34s?  What happens with those?

Should try I guess. Not too many that I can recall.

BG> The reason I ask is that Paul's is just about the only Courier I call
BG> where the initial connect is not always at the maximum link rate,

Maybe, but thats certainly not true of most people calling with USR VEs.

BG> which could indicate that his lines are not 100%

Sure, but again, thats got SFA to do with demanding an immediate retrain
on a V32bis call which is nothing like as demanding as a V34+ call.

Clearly its no surprise that your dont get full speed every
time with V34+, most people dont even if you do with Dave.

BG> (not surprising in a block of units anyway).

I'm not convinced that units are relevant. It is after all
basically copper pairs back to the exchange, the fact that some
of its not in the ground isnt that likely to be that significant.

--- PQWK202