At our POP local to this user we have two PRIs hitting identical PM3s.
Each PRI has its own phone number, so I had him call each number three
times. Then I extracted the Connect-Info from the start and stop
records logged by radius:
User calling PRI #1 on PM3 #1:
Connect-Info = "31200 LAPM/V42BIS"
Connect-Info = "31200 LAPM/V42BIS"
Connect-Info = "31200 LAPM/V42BIS"
Connect-Info = "31200 LAPM/V42BIS"
Connect-Info = "31200 LAPM/V42BIS"
Connect-Info = "31200 LAPM/V42BIS"
SAME user calling PRI #2 on PM3 #2:
Connect-Info = "28800 LAPM/V42BIS"
Connect-Info = "28800 LAPM/V42BIS"
Connect-Info = "4800 LAPM/V42BIS"
Connect-Info = "7200 LAPM/V42BIS"
Connect-Info = "21600 LAPM/V42BIS"
Connect-Info = "21600 LAPM/V42BIS"
Before going further, let me add that this user's modem is a Hayes
Accura K56Flex with no special init string. If the only evidence I
had was what happened when he called the second PRI, I might think his
modem was junk or futzed up, but this user has been using PRI #1 for a
month and has never once connected at less than 31200. I checked the
radius logs.
The difference isn't the PM3s: until last week PRI #1 terminated in
the same T1 port of PM3 #2 that PRI #2 now lands on.
PRI #2 isn't defective. We got 33600 connects, 31200 connects, and
ISDN connects on it by other users the same day. Some 31200 connects
were on the same modem this user was hitting.
What IS different? The call path. PRI #1 originates in the CO local
to the POP and the user, PRI #2 is hauled in from a remote CO.
For PRI #1 and this user the call path is
user <--phone line--> CO-1 <--PRI #1--> PM3
For PRI #2 it's
user <--phone line--> CO-1 <--voice trunk--> CO-2 <--PRI #2--> PM3
The problem is that voice trunk. Just because you have a PRI or CT1
with 8-bit-clean trunk-side attachment doesn't mean the trunks used to
deliver calls to your PRI/CT1 are 8-bit-clean trunks. Quite often
they are not. For this path, in fact, I happen to know that they
virtually never are.
I know good trunks exist for the path, because I can call either PRI
from the other and establish a 64k ISDN connection. Those good trunks
just aren't used for voice calls, which are presumed not to need them.
(Actually, I have evidence a lot of voice calls along the path in the
other direction, CO-2 to CO-1, do use 8-bit-clean trunks.)
I also had this user make three calls to a third PRI hitting a PM3 at
another POP. Same user calling PRI #3:
Connect-Info = "28800 LAPM/V42BIS"
Connect-Info = "28800 LAPM/V42BIS"
Connect-Info = "28800 LAPM/V42BIS"
Connect-Info = "28800 LAPM/V42BIS"
Connect-Info = "31200 LAPM/V42BIS"
Connect-Info = "31200 LAPM/V42BIS"
This is isn't as good - or at least not as certain - as calling the
local PRI, but it's better than than calling PRI #2. When this user
is calling PRI #2, he is calling inward to the largest city around
(Albany, NY). When calling PRI #3 he's calling in the outward
direction. My experience for years - decades actually - has been that
people in a city calling out get better quality than people outside
calling in.
But I'm drifting from my main point: before you jump to conclusions
about your users' modems, know how their calls are routed to you.
Personally I suspect that what's making things difficult for digital
modems is the junk in the robbed signalling and framing bits being
transmitted down PRIs and CT1s with crystal clarity. An analog modem
can pretty much rely on this stuff being filtered out.
One last thing to sneak in: if the list were split up, I'd prefer to
see it split into portmaster-users-on-topic-postings and
portmaster-users-babbling-away.
-- Dick St.Peters, stpeters@NetHeaven.com Gatekeeper, NetHeaven, Saratoga Springs, NY, 1-800-910-6671 (voice) Saratoga/Albany/GlensFalls/NorthCreek/LakePlacid/BlueMountain/Plattsburgh First Internet service based in the 518 area code