[OLPC Networking] Wireless Recommendations (Ricardo Carrano)

[Ricardo Carrano]

Hi Tom.

On Sat, Aug 9, 2008 at 6:50 PM, Tom Mitchell <mitch at niftyegg.com> wrote:
> Ricardo Carrano wrote:
>>
>> On Fri, Aug 8, 2008 at 3:21 PM, Tom Mitchell <mitch at niftyegg.com> wrote:
>>>
>>> Greg Smith wrote:
>>>>
>>>> Hi Ricardo,
>>>>
>>>> Interesting page with some great ideas. A few comments and questions
>>>> after a quick read through.
>>>
>>> Not for the high priority project list:
>>> One connectivity model that could be considered is a
>>> disconnected model.  I think of it as a dial up
>>> on demand uucico model,  uucico for short
>
> .....
>
>>
>> Tom,
>>
>> The subject of DTN (Delay or Disruption Tolerant Networks) are now
>> perceived as increasingly important and I am sure that, in many
>> deployments, technologies emerging from this research and efforts will
>> be very useful.
>>
>> It seems that we have another category of recommendations, or another
>> page, which deals with the school connectivity. So, that would split
>> the deployment recommendations into two categories:
>>
>> - The backhaul (or external) connectivity - the school connecting to
>> the Internet
>> - The wireless (or internal) connectivity - the school wireless
>> infrastructure
>>
>> Comments?
>>
>> Cheers!
>> Ricardo
>
> I suspect that DTN issues will work themselves out. They are core to mobile
> networking futures.
>
> What I was reflecting on is a supplemental backhaul connectivity strategy
> built on existing code and designs.  In the "old days" before wide area
> TCP/IP we moved email, net news, files, sources and more via uucico (unix to
> unix copy in copy out) over telephone line modems when daytime long distance
> calls were expensive and the internet non existent.
>
> Delivery addresses were bang "!" separated routing lists. The telephone
> calls were made on demand, or on a schedule, in the middle of the night, on
> weekends within a time slot window so as to not overload modem banks.   A
> subtle twist to this would be a cross between sneaker net and meshed
> networking where a portable machine could visit collect and deliver files,
> mail, etc. up and down a delivery route on an irregular snail mail like
> schedule.
>
> Like I said not the first thing to work on but something to keep in mind as
> a possible solution for areas with expensive or non existent connectivity.
> It turns out that late versions of uucico work over TCP/IP making
> connectivity transitions almost transparent.
>

And I second your suggestion. [Side note: Let's say I'm not young
enough not to remember uucp. ;-)]

I didn't bring the DTN term to the discussion gratuitously. Actually
the example of the guy that takes a bicycle with a linksys router and
every day rides it through disconnected villages and collect people's
url requests (via email) turns to be one classic scenario of DTN. As
usual, it's just a fancy name for an old problem.

The email seems to be the last remainder of an store-and-forward
application and that's why it probably fits well in this scenario.

I also agree that this may not be a top priority, but I have the same
feeling that there exists the situation where this class of idea (name
is less important) would come in hand.

So, the question is, how can we formalize a deployment recommendation
based on that idea? Would you volunteer? :)

Cheers!
Ricardo