We were extremely skeptical initially as well. But we could find no
rational explanation for it in over a month of looking. I can get the
version numbers of the client side modems. We checked them at the
beginning of the test and again when Rockwell came out in late January.
They were the latest publicly available firmware at that time. Rockwell
brought out the Zoom. It did not perform appreciably better. We ran a bit
of a side test with them using some special software they generated to NOT
fall back to V.34, but then it just timed out most of the time when it
couldnt' get a pcm session.
One of the problems that would seem to relate to your results was the
frequency where the modem gave up and fell back to a V.34 session. If you
look at only PCM sessions, the average would be much higher. But we often
could not get a PCM session and the call fell back to a V.34 session.
Those are definitely counted.
Before the shrill girlish screams of panic, I was very open to thoughts on
this. We couldn't get anything particularly useful from the principles
involved. We have just enough PCM calls to look like it's working, but not
enough to look like its working very well. At this point, I'm very
defensive on the whole topic. If this can break out into screams of rage
based on nothing they know, as in the case of Mr. Denninger, and the gross
accusations of malfeasance from Mr. Sasek, emotions run a wee bit too high
for much to come out of it that is productive. So we'll be very careful
with our sharing. But I am perplexed.
The perspective on modem versions was that we use the latest available.
Rockwell complained about several modem manufacturers being "slow" to
promulgate the latest code. But on arrival claimed we did indeed have the
latest.
jack
----------
> From: M Lyons <lyonsm@netbistro.com>
> To: Dick St.Peters <stpeters@NetHeaven.com>
> Cc: isp-ceo@isp-ceo.com; portmaster-users@livingston.com;
jack.rickard@boardwatch.com
> Subject: Re: [jack.rickard@boardwatch.com: Re: (PM) Re: Nationwide Access
- Please no Dweebs (fwd)]
> Date: Friday, February 20, 1998 3:08 PM
>
>
> On Fri, 20 Feb 1998, Dick St.Peters wrote:
>
> > We see some connections below 40k; we log connections ranging from 32k
> > to 54k. This morning I picked a range of days and analyzed them. Our
> > last 3785 flex connections have averaged 44.56k.
>
> Our average rate for k56flex calls was 43.66kbps over this sample of the
> last 6642 PCM calls to our modest network of 4 PM3's in 2 cities. In our
> location, below 40k calls are certainly not unheard of, as you can see,
> but they are definitely in the minority. If your modem negotiates a PCM
> connection with us, you have about a 92% chance of getting a 42kbps or
> better initial connection.
>
> Of course, these are just the initial connect rates and the modems can
and
> do retrain during a call, so a measure of raw throughput would be
required
> to establish whether the modems initial reported connect rates might be
> consistently too pessimistic or too optimistic, and to what extent the
> retrains themselves affect overall throughput.
>
> For me personally, our results vindicate k56flex as a protocol. While it
> is certainly no better than X2, it is also not substantially worse,
> especially considering that it's implementations are encumbered by
> multi-vendor interoperability issues that do not (currently) affect X2.
>
> My personal feeling is that, to end up with a k56flex average connect
rate
> in the low 30s, there must be some methodological flaw in the Boardwatch
> study which causes it's results to vary from those seen in actual
> practice. Karl Denninger's suggestion that their use of the latest, buggy
> firmware in the client modems caused the skew is quite plausible.. but in
> the world of magazine reviews and marketing pseudo-studies, thems the
> breaks. The PM3 certainly wouldn't be the first product to suffer because
> a magazine reviewer had a run of bad luck.
>
> 39 32000
> 17 34000
> 103 36000
> 255 38000
> 122 40000
> 2175 42000
> 2181 44000
> 1211 46000
> 94 48000
> 405 50000
> 40 52000
>
> -m
-
To unsubscribe, email 'majordomo@livingston.com' with
'unsubscribe portmaster-users' in the body of the message.
Searchable list archive: <URL:http://www.livingston.com/Tech/archive/>