[Cu-wireless] an article on an 802.11 rooftop network

[stephane_alnet at ureach.com]

Dave,

> Sorry, I think I transposed the S and V: it is DSDV that they use.

Oh, I was just pointing out that from what I understood of the MIT paper,
they did most if not all of the testing using static "routes". The paper
effectively mentions DSDV in the intro; I admit I went a bit rapidly
through the last part of the document and jumped to the part that said
"conclusion", so I may have missed the section where they went into DSDV.
:-\

> Direct Sequence Distance Vector, I think.  It is in the book Ad Hoc
> Networking, isn't it?

It is, with Perkins noting it's mostly there for historical purposes and
that it was designed "quite before [they] thought about "on-demand", so
DSDV looks primitive by comparison with modern protocols." [p.23]
(You can bet I skipped that chapter.)

> I forget what other reasons that guy gave that 802.11b is not appropriate
> for ad hoc networks.

The two main reasons I've seen (and can remember) are (a) RTS/CTS is a
pain and (b) 802.11 doesn't behave well with asymmetrical links (which is
also a problem for AODV at L3).

I had read about RTS/CTS somewhere else, but the MIT paper is the first
paper (I've seen) to point out the issue with asymmetry at L2 (and the
fact that asymmetry is quite common). Actually this is the first paper
I've seen that goes this far in analyzing the issues we are going to face
with 802.11b -- reading just the conclusion of the paper can give one a
good overall idea. (The DSR paper had some testing, with the three cars on
a loop between the two buildings, but that was 900MHz technology; all the
other papers/protocols seem to use simulations to "prove" they are
better.)

> I think that there are some challenges, and 802.11b is certainly not the
> best (I think 900MHz is a more suitable band, and there's gotta be a
> better MAC protocol), but I think 802.11b is "good enough."

And it's cheap too. And it has very good cross-platform support. And it's
available off-the-shelf. -- No point intended; I was just sharing my
"reading comments". :-]

> There are different MAC protocols than 802.11 [...].

That's where the DoD went if I understand properly. Until such point we
can trick the firmware (if ever), we'll have to live with some of 802.11
shortcomings. (Knowing what they are may help us until such time.)

But again, I haven't been impressed with what I've seen so far as far as
layer 3 routing is concerned; it sounds like "scalability" in the ad-hoc
world means "50 nodes (on paper)". As far as I can tell from my limited
reading on the subject, it doesn't seem anybody has actually tested any of
this stuff in the real world with a proper "scale" that would match what
we are shooting for. Which is something you said a couple weeks ago. ;)

S.