DSL service [was Re:TLUG Mail list archives?]

[D. Hugh Redelmeier]

| From:  <phiscock-g851W1bGYuGnS0EtXVNi6w at public.gmane.org>

| Here (below) is a typical result from pinging a website. I get a similar
| result irrespective of the website. It pings fine for a while and then the
| site becomes unreachable.

Yuck.

A more interesting test is with mtr.  It is a traceroute but with more
parallelism.  Of course there must be something better by now but I'm
don't follow this stuff closely.

Anyway, that would tell you what links are problematic and some of the
nature of the problem (100% lost packets?  50% lost packets?  slow
packets?).

| This shouldn't be a distance problem, we're less than 1km from the CO -
| and the Bell tech who was here said I should be getting something like
| 5MB/sec at this location.
| 
| I wonder if the Sympatico service is any better than the other ISPs? You
| could then see this as a strategy to drive customers into Bell's embrace.

Sympatico would not be my first (or second or third) choice.  You get
treated like an AOLer.  I like being able to have a static address,
running my own mail server, not having my traffic shaped (i.e.
bittorrent throttled) (that's really a Rogers disease at the moment,
but Sympatico might catch it).

| Anyway, I'm  checking out Rogers: it's about $10 more per month, but
| reliable fast service would be worth it.

I don't recommend Rogers unless you don't mind the afformentioned
traffic shaping and port blocking.  Oh and the terms of service that
ban servers (a thing not well-defined on the internet and thus giving
them more arbitrary power).

I have Rogers and ADSL.  Rogers is fast and mostly reliable, but treat
me as a "consumer".  With ADSL, I have a bad wire to the CO, but the
rest is pretty good.

| 64 bytes from 62.41.81.27: icmp_seq=542 ttl=54 time=61.8 ms

This latency seems high.  With mtr or tracerout output you could tell
where the bottleneck is.

When I first got ADSL (with a now-defunct but small and cool ISP) I
had consistently high latency (about 60ms to get to the first hop).
The cause was interleaving of each IP packet when mapped into ATM
packets for ADSL transmission.  My ISP didn't even know of such a
thing.  It was finally explained to me by a Bell Labs guy!  Just now I
found this stuff using google:

  http://www.wlug.org.nz/Interleaving
  http://cable-dsl.home.att.net/#Interleaved
  http://www.faqs.org/docs/Linux-HOWTO/DSL-HOWTO.html#AEN680

| >From 192.168.7.1 icmp_seq=570 Destination Net Unreachable

I am guessing that 192.168.7.1 is your local router's address, on your
side of the ADSL link (since it is an RFC 1918 address).  As such,
this implies that something unsubtle is wrong.  Perhaps that the link
is down.  Perhaps your PPPoE implementation is not logging in
reliably.

On my ADSL system, the modem is only a modem.  The PPPoE crap is
implemented on my router which is a Linux box.  Actually, lots of
routers are Linux boxes.  I mean that my router is actually a PC
running Red Hat Linux that I have configured to be a router.  This has
several advantages (I can log into my router and figure out what is
going on when it isn't working, very flexible, good logging) and
several disadvantages (it isn't a turnkey appliance, it burns more
power than a dedicated consumer broadband router).  When I got ADSL
there were no consumer broadband routers that implemented PPPoE AFAIK.
The reason that I'm mentioning this is that my debugging procedures
might be different from yours because I have different tools
available.  Like tcpdump on the raw PPPoE stream, for example.
--
The Toronto Linux Users Group.      Meetings: http://gtalug.org/
TLUG requests: Linux topics, No HTML, wrap text below 80 columns
How to UNSUBSCRIBE: http://gtalug.org/wiki/Mailing_lists