[Gllug] Internet connection through mobile phone

general_email at technicalbloke.com general_email at technicalbloke.com
Sat Dec 12 10:22:45 UTC 2009

Jason Clifford wrote:
> On Sat, 2009-12-12 at 08:31 +0000, general_email at technicalbloke.com
> wrote:
>> OP, as I mentioned in an earlier post I don't think mobile broadband is
>> all that bad if your needs are modest (I don't think LUG members
>> generally have modest demands of their technology!) 
> members have referred to serious problems with connections dropping, the
> speed being practically non-existent and other problems on mobile
> services. I've experienced all of these. These problems are not to do
> with excessive demands but rather the simple fact that mobile broadband
> services are just not fit for any kind of regular use - even just basic
> web browsing.

Well that's not my experience, other than low speed in some parts of my
house it seems fine. I was doing 'basic web browsing' on it yesterday
with no problems. Your 'simple fact' is a just a fairly blunt conjecture
and I refer you to your argument below: "the vast majority of people
simply don't have problems with the service. If they did ISPs would not
be able to operate as the margins are so low.". Which is it?

>> and it's certainly
>> cheap to try so I'd encourage you to give it a go. £120 + £15 /month
>> line rental = £300 for your first year of landline ownership and that's
>> before your broadband bill!
> Yes activation can be expensive but £15/month line rental is a fantasy.
> The usual price is less than £12/month. A decent broadband service on
> top of that will cost more. How much more will depend upon how good a
> service you want.

It's not a fantasy if you don't pay by direct debit and don't cancel
paper bills. Not only that but it will be going up an extra 50p/month
next year to subsidise rural connections. It makes little difference
anyway, being more precise and using your figure gives £122.50 + 12*12 =
£266.50 a year i.e. still an awful lot more expensive than 3G for
moderate use and that's not including the cost of a broadband package!

BTW, price and quality aren't directly correlated at all. The service I
have now, the uncapped "Be Value" is only £13.50 a month and is by far
the best broadband experience I've had in London. The BT connection I
had to endure before cost twice as much and was utterly shit.

>> If you do go with fixed line broadband it's hard to recommend anyone as
>> being very good, I see frequent problems with pretty much every
>> provider.
> More often than not the problems are specific to customer's having
> unrealistic expectations (do read the "up to" as meaning just that!) or
> to customer wiring or equipment but it's certainly true that problems
> can occur. When problems do occur with a fixed line broadband service
> it's a lot easier to determine where the problem is than on a mobile
> service.

That may be true but ease of fault finding is hardly worth over £250 for
most domestic users. I have to say I consider this use of "up to" in
marketing thoroughly dishonest anyway, I have come across providers
advertising upto 20 Mb that can't consistently deliver one. I'd like to
see them forced to advertise their average speed as monitored by an
independent 3rd party, that would quickly bust the myth that paying
extra for BT gets you fast reliable service.

>> It's hard to find any objective data on the reliability of various
>> broadband providers but I suspect they are all fairly equally flaky,
>> they certainly seem to be. This would make sense as the vast majority of
>> the electronics and cables involved in anyone's ADSL connection are
>> BT/Openzones, not the broadband companies.
> And the vast majority of people simply don't have problems with the
> service. If they did ISPs would not be able to operate as the margins
> are so low.

Did I say that they did!? I'm a repair man so I see a lot of setups that
are struggling, not so many that are working fine. What I am saying is
that these problems seem to be fairly evenly spread across service
providers. I have come across problems on everybody's networks and no
one provider stands out as being particularly reliable or unreliable.
This is one of the reasons I object to BT's repeated use of the words
"Fast and Reliable" in their advertising, I have experienced a boatload
of slow flaky BT connections and yes, I have tried swapping out the
routers and connecting them directly to the master sockets. Even
factoring in their ubiquity BT provide a pretty dismal service IMHO and
they have some remarkable chutzpah to market it as "Fast and Reliable"

On a side note those TV adverts of theirs make me want to bludgeon my
telly and everything within a 50 yard radius to a Planckian thickness.

>> With more objective data gathering in mind I recently wrote a basic
>> reliability monitor in Python. I occasionally get a client who seems /
>> claims to be experiencing above average dropouts so I figured this would
>> be a good way to test. It periodically pings various public servers and
>> logs the results. Turns out my connection isn't quite as rock solid as I
>> had thought it was! Anyone fancy trying it out on their connection /
>> helping me develop it?
> You're not doing a reliable test. Ping to remote sites cannot be relied
> upon to provide an accurate indication of how good or bad your
> connection is as ICMP is often blocked or marked for low priority in the
> event of any congestion anywhere on the network and you cannot know if
> that is being done or where it is being done when measuring connection
> to a remote site.

Actually, although it's still very early days, it seems good enough to
indicate bad service. I have profiled the worlds top 25000 websites for
over a week and weeded out any that do not respond very promptly to
ICMP. Those that made it through I then allow a generous time out.
Naturally a failure is only considered an outage once a threshold of
consecutive failures has been exceeded and any servers that begin to
fail a disproportionate amount of time will be removed form the list. I
have tested on several Windows and Linux computers and ICMP failures
correspond very well with failures of web service, in future versions I
plan to add an option to test http periodically too.

> To test a broadband connection you need to be using LCP to test between
> the end point and the ISP network. That's the only way to measure the
> broadband connection.

Balls. Just last week it identified I was getting periodic spotty DNS
service, I contacted my ISP and they identified out of date DNS settings
on my router (supposedly managed and updated by them). They corrected
these by hand and I have had considerably better service ever since. I
see no reason why, given a known good computer and router the above
method could not generate quite useful results in most domestic/SoHo

>  It's also
> a waste of resources in almost every case.

<3MB memory use, unmeasurable processor use and with 350 bytes * every 5
seconds = 4.3Mb per day, spread across > 20,000 servers its hardly going
to cause the internet to grind to a halt is it?

Gllug mailing list  -  Gllug at gllug.org.uk

More information about the GLLUG mailing list