Re: [Sugar-devel] [Design] Ad-hoc networks - New Icons

2009-08-21 Thread Bobby Powers
On Wed, Aug 19, 2009 at 11:59 AM, Sameer Vermasve...@sfsu.edu wrote:
 On Wed, Aug 19, 2009 at 5:38 AM, Benjamin M.
 Schwartzbmsch...@fas.harvard.edu wrote:
 Sameer Verma wrote:
 Hi Ben,

 So, you were referring to NM's inability to handle switching to hostap
 (making the wireless card act as an AP)? http://hostap.epitest.fi/

 I'm aware of hostapd.  In fact, I'm running it right now on an Athlon box
 in the living room, which acts as my apartment's access point.

 Yes, I did the same for many years...quite a learning experience. We
 used to run our college's network off a 133MHz Pentium laptop on RH 6
 :-)

 I'm merely
 noting that hostapd (or equivalent) is not yet available via
 NetworkManager, so implementing AP mode in Sugar would require a
 significant restructuring of the networking code.

 Does the driver for Marvell chipset on the XO support hostapd (outside
 of NM of course)?

I don't believe so.  I remember hearing that there's not enough room
in Flash on the XO1's Marvell chip for firmware that does both regular
client/mesh as well as hostap.

Bobby

 Sameer

 --Ben


 ___
 Sugar-devel mailing list
 Sugar-devel@lists.sugarlabs.org
 http://lists.sugarlabs.org/listinfo/sugar-devel

___
Sugar-devel mailing list
Sugar-devel@lists.sugarlabs.org
http://lists.sugarlabs.org/listinfo/sugar-devel


Re: [Sugar-devel] [Design] Ad-hoc networks - New Icons

2009-08-19 Thread Benjamin M. Schwartz
Sameer Verma wrote:
 Hi Ben,
 
 So, you were referring to NM's inability to handle switching to hostap
 (making the wireless card act as an AP)? http://hostap.epitest.fi/

I'm aware of hostapd.  In fact, I'm running it right now on an Athlon box
in the living room, which acts as my apartment's access point.  I'm merely
noting that hostapd (or equivalent) is not yet available via
NetworkManager, so implementing AP mode in Sugar would require a
significant restructuring of the networking code.

--Ben



signature.asc
Description: OpenPGP digital signature
___
Sugar-devel mailing list
Sugar-devel@lists.sugarlabs.org
http://lists.sugarlabs.org/listinfo/sugar-devel


Re: [Sugar-devel] [Design] Ad-hoc networks - New Icons

2009-08-17 Thread Simon Schampijer
Let me summarize this thread:

a) User Point of view
Mesh: created automatically, small networks

Ad-Hoc: user created, very small networks, internet connection sharing

b) Technically
Mesh: not range limited (package forwarding), no creator principle

Ad-Hoc: range limited, best 2 people to avoid cases where C can see A 
and B, but A and B are unaware of each others presence, still present 
when the creator leaves

Which icon to pick:
The ad-hoc network is not really well presented with a badged mesh icon 
as it may suggest non existing properties. As well a badged AP icon is 
not optimal, as the network can persist if the creator leaves. Though, 
if the ad-hoc network is only for very small networks I wonder if we 
would see the case, where the creator leaves before the others, often. 
The quality of an infrastructure is given, in that the creator provides 
his guests with internet facilities if present.

I would vote for a badged AP icon (if it is distinguishable enough), or 
a complete new icon (should be based on a circle, as the wireless device 
icon is a circle as well).

Regards,
Simon

___
Sugar-devel mailing list
Sugar-devel@lists.sugarlabs.org
http://lists.sugarlabs.org/listinfo/sugar-devel


Re: [Sugar-devel] [Design] Ad-hoc networks - New Icons

2009-08-17 Thread Benjamin M. Schwartz
Daniel Drake wrote:
 2009/8/11 Gary C Martin g...@garycmartin.com:
 H. Are you sure this is an accurate statement? I was under the
 impression that mesh forwarding support had been removed/disabled from OLPCs
 implementation a long time ago, since soon after the Mongolia deployment.
 Mesh was killing the wireless spectrum with all the attempted packet
 retransmissions. It is really only 'mesh' in name, all devices have to be in
 range of each other to collaborate.
 
 Yes, forwarding still happens.

It does, in mesh mode.  Mesh mode is still built into the default
software, but OLPC encourages deployments to install standard wireless
access points in the schools, instead of using the mesh there.

What OLPC has deprecated is Mesh Portal (MPP) mode, in which XOs connect
to both an access point _and_ to a mesh network, on the same channel, and
route packets between them.  This setup was designed with the idea that it
would allow a small number of access points to cover a large area.
Unfortunately, the overhead of forwarding packets through the mesh has
thus far exceeded the gain in covered area.

--Ben



signature.asc
Description: OpenPGP digital signature
___
Sugar-devel mailing list
Sugar-devel@lists.sugarlabs.org
http://lists.sugarlabs.org/listinfo/sugar-devel


Re: [Sugar-devel] [Design] Ad-hoc networks - New Icons

2009-08-17 Thread Benjamin M. Schwartz
Peter Robinson wrote:
 Hmm... so if other people join that network, and then I leave, does the
 network not persist?  I haven't experimented with this yet.
 yes, it will persist
 
 Really? My understanding of the ad-hoc network is that it basically
 puts the user that creates it wifi card into ad-hoc AP mode and that
 machine basically acts as a AP and link local IPs are used.

There's no such thing as ad hoc AP mode; ad-hoc and AP are opposites.
In an ad-hoc network, all traffic is sent directly from one participant to
another.  In AP (a.k.a infrastructure, master, host AP) mode, messages are
sent to the AP, which forwards them to the recipient.

An ad hoc network has no owner, despite the proposed portrayal in the
UI.  Personally, I think that if we're going to use such a display, we
should actually implement it over infrastructure mode, with the network
owner running in host mode.  This should provide much less surprising
behavior.  In particular, in infrastructure mode, there is no mutual
routability problem.  Anyone who can reach the host is connected, and can
route to everyone else.

Both ad hoc mode and AP mode require some level of driver support, and the
maturity of this support varies between drivers.  Also, I believe
NetworkManager still has no concept of AP mode, so it may not yet be
possible to include AP mode in Sugar.  Perhaps we can change the backend
in a future Sugar release, once NetworkManager is ready.

--Ben



signature.asc
Description: OpenPGP digital signature
___
Sugar-devel mailing list
Sugar-devel@lists.sugarlabs.org
http://lists.sugarlabs.org/listinfo/sugar-devel


Re: [Sugar-devel] [Design] Ad-hoc networks - New Icons

2009-08-17 Thread Mathieu Bridon (bochecha)
 Also, I believe
 NetworkManager still has no concept of AP mode

Do you mean that with NM you can't become an AP ?

Something like that:
http://magazine.redhat.com/2008/10/16/video-fedora-10-connection-sharing/
?

Or did I misunderstand what you meant ?


--

Mathieu Bridon (bochecha)
___
Sugar-devel mailing list
Sugar-devel@lists.sugarlabs.org
http://lists.sugarlabs.org/listinfo/sugar-devel


Re: [Sugar-devel] [Design] Ad-hoc networks - New Icons

2009-08-17 Thread Benjamin M. Schwartz
Mathieu Bridon (bochecha) wrote:
  Also, I believe
 NetworkManager still has no concept of AP mode
 
 Do you mean that with NM you can't become an AP ?

Yes.  That video does _not_ show the connection-sharing computer becoming
an Access Point.  It shows it becoming a node on an ad hoc network.  This
means that, unlike an Access Point, it will not forward packets between
other computers on the network, thus creating the mutual routability
problems that we have been describing.

--Ben



signature.asc
Description: OpenPGP digital signature
___
Sugar-devel mailing list
Sugar-devel@lists.sugarlabs.org
http://lists.sugarlabs.org/listinfo/sugar-devel


Re: [Sugar-devel] [Design] Ad-hoc networks - New Icons

2009-08-11 Thread Daniel Drake
2009/8/11 Simon Schampijer si...@schampijer.de:
 I think it would help, to have a new icon for the ad-hoc network to
 distinguish them. Could be a badged wireless network one? Or is the mesh
 icon appropriate? Or something completely new?

I think new icons would be best, to distinguish from the mesh. I think
we can expect mesh support again soon ;)

Daniel
___
Sugar-devel mailing list
Sugar-devel@lists.sugarlabs.org
http://lists.sugarlabs.org/listinfo/sugar-devel


Re: [Sugar-devel] [Design] Ad-hoc networks - New Icons

2009-08-11 Thread Simon Schampijer
On 08/11/2009 11:50 AM, Daniel Drake wrote:
 2009/8/11 Simon Schampijersi...@schampijer.de:
 I think it would help, to have a new icon for the ad-hoc network to
 distinguish them. Could be a badged wireless network one? Or is the mesh
 icon appropriate? Or something completely new?

 I think new icons would be best, to distinguish from the mesh. I think
 we can expect mesh support again soon ;)

 From the user POV they are the same I guess. A local network, that does 
not need any infrastructure.

Though, the mesh on the XO is handled automatically, the ad-hoc network 
requires user interaction to create it. I wonder if we ever will see a 
user using both (not at the same time) on the same machine. To think 
about the visual clash, at least.

Regards,
Simon
___
Sugar-devel mailing list
Sugar-devel@lists.sugarlabs.org
http://lists.sugarlabs.org/listinfo/sugar-devel


Re: [Sugar-devel] [Design] Ad-hoc networks - New Icons

2009-08-11 Thread Gary C Martin
On 11 Aug 2009, at 11:21, Simon Schampijer wrote:

 On 08/11/2009 11:50 AM, Daniel Drake wrote:
 2009/8/11 Simon Schampijersi...@schampijer.de:
 I think it would help, to have a new icon for the ad-hoc network to
 distinguish them. Could be a badged wireless network one? Or is  
 the mesh
 icon appropriate? Or something completely new?

 I think new icons would be best, to distinguish from the mesh. I  
 think
 we can expect mesh support again soon ;)

 From the user POV they are the same I guess. A local network, that  
 does
 not need any infrastructure.

 Though, the mesh on the XO is handled automatically, the ad-hoc  
 network
 requires user interaction to create it. I wonder if we ever will see a
 user using both (not at the same time) on the same machine. To think
 about the visual clash, at least.

I do wonder if the ad-hoc network should actually be being auto  
magically created, if the owner is not associated with an available  
AP, much like the mesh was. We would agree a standard network name (as  
did olpc-mesh), that way there is a minimum of user required  
interaction and any Sugar users in range would auto connect to the  
same network for collaboration.

Regards,
--Gary

P.S. Apologies, I've not yet seen any of the current UI solution from  
Tomeu other than some screen shots, as most of my testing is in a VM  
that does not emulate wireless networks. Is this new feature exposed  
in the device frame, with an always visible ad-hoc device icon?

___
Sugar-devel mailing list
Sugar-devel@lists.sugarlabs.org
http://lists.sugarlabs.org/listinfo/sugar-devel


Re: [Sugar-devel] [Design] Ad-hoc networks - New Icons

2009-08-11 Thread Peter Robinson
 2009/8/11 Simon Schampijersi...@schampijer.de:
 I think it would help, to have a new icon for the ad-hoc network to
 distinguish them. Could be a badged wireless network one? Or is
 the mesh
 icon appropriate? Or something completely new?

 I think new icons would be best, to distinguish from the mesh. I
 think
 we can expect mesh support again soon ;)

 From the user POV they are the same I guess. A local network, that
 does
 not need any infrastructure.

 Though, the mesh on the XO is handled automatically, the ad-hoc
 network
 requires user interaction to create it. I wonder if we ever will see a
 user using both (not at the same time) on the same machine. To think
 about the visual clash, at least.

 I do wonder if the ad-hoc network should actually be being auto
 magically created, if the owner is not associated with an available
 AP, much like the mesh was. We would agree a standard network name (as
 did olpc-mesh), that way there is a minimum of user required
 interaction and any Sugar users in range would auto connect to the
 same network for collaboration.

The problem with that would be that you'd have a number of devices
suddenly sharing the same network. If your in a group of users unlike
in the mesh environment only one person/device would create this.

Peter
___
Sugar-devel mailing list
Sugar-devel@lists.sugarlabs.org
http://lists.sugarlabs.org/listinfo/sugar-devel


Re: [Sugar-devel] [Design] Ad-hoc networks - New Icons

2009-08-11 Thread Gary C Martin
On 11 Aug 2009, at 12:08, Peter Robinson wrote:

 2009/8/11 Simon Schampijersi...@schampijer.de:
 I think it would help, to have a new icon for the ad-hoc network  
 to
 distinguish them. Could be a badged wireless network one? Or is
 the mesh
 icon appropriate? Or something completely new?

 I think new icons would be best, to distinguish from the mesh. I
 think
 we can expect mesh support again soon ;)

 From the user POV they are the same I guess. A local network, that
 does
 not need any infrastructure.

 Though, the mesh on the XO is handled automatically, the ad-hoc
 network
 requires user interaction to create it. I wonder if we ever will  
 see a
 user using both (not at the same time) on the same machine. To think
 about the visual clash, at least.

 I do wonder if the ad-hoc network should actually be being auto
 magically created, if the owner is not associated with an available
 AP, much like the mesh was. We would agree a standard network name  
 (as
 did olpc-mesh), that way there is a minimum of user required
 interaction and any Sugar users in range would auto connect to the
 same network for collaboration.

 The problem with that would be that you'd have a number of devices
 suddenly sharing the same network. If your in a group of users unlike
 in the mesh environment only one person/device would create this.


Apologies for being technically naive;

1). I think if using the same SSID and channel number in ad-hoc mode,  
devices will work together. There's no security, authentication, the  
wireless NICS are all just randomly broadcasting and listening.

Tomeu: If has made it into one of the XO builds, I can run tests next  
week (3 XOs + 1 Mac).

2). Alternatively if I'm wrong about 1, how about a behaviour that  
auto create a default ad-hoc network if it's not visible already, and  
joins one if it is? If the creator goes away/offline, the network  
obviously fails and one of the other clients creates it again (after  
short random delay), and the rest re-auto join.

Thinking about the benefits of a manual ad-hoc process; it does allow  
a (technically aware) teacher to create a named wireless network on  
their machine for their class to join, thereby helping isolate  
different working groups of students. Perhaps also when a class is  
split into working groups, the team leader of each could be instructed  
to create a named ad-hoc network for the rest of their group to use  
(though not sure how able our demographic would be for such an  
operation, probably 9-12 year olds would be capable).

Think I'd still much prefer the ad-hoc as mesh-like auto set-up  
behaviour, it's better for out target demographic, reduces UI, and  
lets collaboration 'just work' when no AP is in use.

Regards,
--Gary

___
Sugar-devel mailing list
Sugar-devel@lists.sugarlabs.org
http://lists.sugarlabs.org/listinfo/sugar-devel


Re: [Sugar-devel] [Design] Ad-hoc networks - New Icons

2009-08-11 Thread Martin Dengler
On Tue, Aug 11, 2009 at 01:35:15PM +0100, Gary C Martin wrote:
 Tomeu: If has made it into one of the XO builds, I can run tests next  
 week (3 XOs + 1 Mac).

It's in my SoaS-on-XO-1 builds (and other SoaS builds, I believe).
 
 Regards,
 --Gary

Martin


pgpbBmIo3eeth.pgp
Description: PGP signature
___
Sugar-devel mailing list
Sugar-devel@lists.sugarlabs.org
http://lists.sugarlabs.org/listinfo/sugar-devel


Re: [Sugar-devel] [Design] Ad-hoc networks - New Icons

2009-08-11 Thread Eben Eliason
On Tue, Aug 11, 2009 at 6:21 AM, Simon Schampijersi...@schampijer.de wrote:
 On 08/11/2009 11:50 AM, Daniel Drake wrote:

 2009/8/11 Simon Schampijersi...@schampijer.de:

 I think it would help, to have a new icon for the ad-hoc network to
 distinguish them. Could be a badged wireless network one? Or is the mesh
 icon appropriate? Or something completely new?

 I think new icons would be best, to distinguish from the mesh. I think
 we can expect mesh support again soon ;)

 From the user POV they are the same I guess. A local network, that does not
 need any infrastructure.

 Though, the mesh on the XO is handled automatically, the ad-hoc network
 requires user interaction to create it. I wonder if we ever will see a user
 using both (not at the same time) on the same machine. To think about the
 visual clash, at least.

Perhaps we could use the mesh icon with a little XO badge, to indicate
that it's functionally similar to the real mesh, but enabled by a
specific XO. Thinking about this now, it might be the case that Tomeu
had built this functionality as an extension of the wireless network
device in the Frame; Should it be an extension of the mesh device
instead, based on its perceived similarities to that feature more than
an AP?

Eben
___
Sugar-devel mailing list
Sugar-devel@lists.sugarlabs.org
http://lists.sugarlabs.org/listinfo/sugar-devel


Re: [Sugar-devel] [Design] Ad-hoc networks - New Icons

2009-08-11 Thread Daniel Drake
2009/8/11 Simon Schampijer si...@schampijer.de:
 From the user POV they are the same I guess. A local network, that does not
 need any infrastructure.

I disagree. The mesh connections are automatic, and the presence of
them does not indicate the presence of another computer like an ad-hoc
network would do. Also, they do not have the principle of ownership
that sugar places on ad-hoc networks.
The behavioural properties of the networks (including the likelihood
of communication) are different because there is no forwarding of
frames. Also, neighbouring but sleeping laptops will not forward
frames on an ad-hoc network.

Daniel
___
Sugar-devel mailing list
Sugar-devel@lists.sugarlabs.org
http://lists.sugarlabs.org/listinfo/sugar-devel


Re: [Sugar-devel] [Design] Ad-hoc networks - New Icons

2009-08-11 Thread Simon Schampijer
On 08/11/2009 02:35 PM, Gary C Martin wrote:
 On 11 Aug 2009, at 12:08, Peter Robinson wrote:

 2009/8/11 Simon Schampijersi...@schampijer.de:
 I think it would help, to have a new icon for the ad-hoc network to
 distinguish them. Could be a badged wireless network one? Or is
 the mesh
 icon appropriate? Or something completely new?

 I think new icons would be best, to distinguish from the mesh. I
 think
 we can expect mesh support again soon ;)

 From the user POV they are the same I guess. A local network, that
 does
 not need any infrastructure.

 Though, the mesh on the XO is handled automatically, the ad-hoc
 network
 requires user interaction to create it. I wonder if we ever will see a
 user using both (not at the same time) on the same machine. To think
 about the visual clash, at least.

 I do wonder if the ad-hoc network should actually be being auto
 magically created, if the owner is not associated with an available
 AP, much like the mesh was. We would agree a standard network name (as
 did olpc-mesh), that way there is a minimum of user required
 interaction and any Sugar users in range would auto connect to the
 same network for collaboration.

 The problem with that would be that you'd have a number of devices
 suddenly sharing the same network. If your in a group of users unlike
 in the mesh environment only one person/device would create this.


 Apologies for being technically naive;

 1). I think if using the same SSID and channel number in ad-hoc mode,
 devices will work together. There's no security, authentication, the
 wireless NICS are all just randomly broadcasting and listening.

 Tomeu: If has made it into one of the XO builds, I can run tests next
 week (3 XOs + 1 Mac).

 2). Alternatively if I'm wrong about 1, how about a behaviour that auto
 create a default ad-hoc network if it's not visible already, and joins
 one if it is? If the creator goes away/offline, the network obviously
 fails and one of the other clients creates it again (after short random
 delay), and the rest re-auto join.

 Thinking about the benefits of a manual ad-hoc process; it does allow a
 (technically aware) teacher to create a named wireless network on their
 machine for their class to join, thereby helping isolate different
 working groups of students. Perhaps also when a class is split into
 working groups, the team leader of each could be instructed to create a
 named ad-hoc network for the rest of their group to use (though not sure
 how able our demographic would be for such an operation, probably 9-12
 year olds would be capable).

I like that group work. I always thought of the ad-hoc network being 
something a group of kids could connect to for a group work. Not 
something the whole school would connect to.

 Think I'd still much prefer the ad-hoc as mesh-like auto set-up
 behaviour, it's better for out target demographic, reduces UI, and lets
 collaboration 'just work' when no AP is in use.

I wonder if the ad-hoc network will scale up to that number of users :/ 
Though I am not an expert in this area. Maybe Daniel has some more 
insights on this topic.

Regards,
Simon

___
Sugar-devel mailing list
Sugar-devel@lists.sugarlabs.org
http://lists.sugarlabs.org/listinfo/sugar-devel


Re: [Sugar-devel] [Design] Ad-hoc networks - New Icons

2009-08-11 Thread Simon Schampijer
On 08/11/2009 03:49 PM, Eben Eliason wrote:
 On Tue, Aug 11, 2009 at 6:21 AM, Simon Schampijersi...@schampijer.de  wrote:
 On 08/11/2009 11:50 AM, Daniel Drake wrote:
 2009/8/11 Simon Schampijersi...@schampijer.de:
 I think it would help, to have a new icon for the ad-hoc network to
 distinguish them. Could be a badged wireless network one? Or is the mesh
 icon appropriate? Or something completely new?
 I think new icons would be best, to distinguish from the mesh. I think
 we can expect mesh support again soon ;)
  From the user POV they are the same I guess. A local network, that does not
 need any infrastructure.

 Though, the mesh on the XO is handled automatically, the ad-hoc network
 requires user interaction to create it. I wonder if we ever will see a user
 using both (not at the same time) on the same machine. To think about the
 visual clash, at least.

 Perhaps we could use the mesh icon with a little XO badge, to indicate
 that it's functionally similar to the real mesh, but enabled by a
 specific XO. Thinking about this now, it might be the case that Tomeu
 had built this functionality as an extension of the wireless network
 device in the Frame; Should it be an extension of the mesh device
 instead, based on its perceived similarities to that feature more than
 an AP?

 Eben

Yes, as of today, it is an extension of the wireless network frame 
device. 
http://1.bp.blogspot.com/_OoKpv4QinxI/SiFiDO2RsAI/ACU/go8n8S6rrE0/s1600-h/soas-create.png

 From the similarities, I agree, a badged mesh icon would work well to 
demonstrate that.

Another question is the behavior: Gary and some others were wondering if 
we should fallback to an adhoc network automatically, if we are not 
connected to an AP.

Regards,
Simon
___
Sugar-devel mailing list
Sugar-devel@lists.sugarlabs.org
http://lists.sugarlabs.org/listinfo/sugar-devel


Re: [Sugar-devel] [Design] Ad-hoc networks - New Icons

2009-08-11 Thread Benjamin M. Schwartz
Daniel Drake wrote:
 2009/8/11 Simon Schampijer si...@schampijer.de:
 From the user POV they are the same I guess. A local network, that does not
 need any infrastructure.
 
 I disagree. The mesh connections are automatic, and the presence of
 them does not indicate the presence of another computer like an ad-hoc
 network would do. Also, they do not have the principle of ownership
 that sugar places on ad-hoc networks.

I am becoming very confused.  Why does Sugar place an ownership concept on
an ad-hoc network?

An ad-hoc network isn't owned by anyone.  It's just spectrum.

--Ben



signature.asc
Description: OpenPGP digital signature
___
Sugar-devel mailing list
Sugar-devel@lists.sugarlabs.org
http://lists.sugarlabs.org/listinfo/sugar-devel


Re: [Sugar-devel] [Design] Ad-hoc networks - New Icons

2009-08-11 Thread Daniel Drake
2009/8/11 Simon Schampijer si...@schampijer.de:
 I wonder if the ad-hoc network will scale up to that number of users :/
 Though I am not an expert in this area. Maybe Daniel has some more
 insights on this topic.

Yes. Ad-hoc networks do not scale at all because they are range limited.

They also act quite odd in terms of networks because you frequently
have situations where C can see A and B, but A and B are unaware of
each others presence. This will lead to funky situations where C could
share an activity, A and B could join without problems, but any
actions by A would not be seen by B and vice-versa.

The basic philosophy behind ad-hoc networks is shout unless someone
else is shouting which also ends up causing various network splices
and merges thanks to the poor quality of radios in the world, and
interference.

In short, ad-hoc networks are not reliable (especially when there are
more than 2 participants) and are very quirky. It is a nice feature to
have in sugar but it is not something that should be created
automatically since in many cases they will just waste airtime for
reliable networks and cause headaches for users.

Automatic infrastructure-less networks are still very desirable of
course, and the best solution that I know of for that would be
open80211s which needs work done lower in the stack before it can be
directly usable by sugar.

and we're miles off topic on a thread about icons.. sigh
Daniel
___
Sugar-devel mailing list
Sugar-devel@lists.sugarlabs.org
http://lists.sugarlabs.org/listinfo/sugar-devel


Re: [Sugar-devel] [Design] Ad-hoc networks - New Icons

2009-08-11 Thread Daniel Drake
2009/8/11 Benjamin M. Schwartz bmsch...@fas.harvard.edu:
 I am becoming very confused.  Why does Sugar place an ownership concept on
 an ad-hoc network?

Just in principle. Have you tried it?
Create a network and it will be called Ben's network
I'd be relatively confident that I could find Ben on that network.

Daniel
___
Sugar-devel mailing list
Sugar-devel@lists.sugarlabs.org
http://lists.sugarlabs.org/listinfo/sugar-devel


Re: [Sugar-devel] [Design] Ad-hoc networks - New Icons

2009-08-11 Thread Benjamin M. Schwartz
Daniel Drake wrote:
 2009/8/11 Benjamin M. Schwartz bmsch...@fas.harvard.edu:
 I am becoming very confused.  Why does Sugar place an ownership concept on
 an ad-hoc network?
 
 Just in principle. Have you tried it?
 Create a network and it will be called Ben's network
 I'd be relatively confident that I could find Ben on that network.

Hmm... so if other people join that network, and then I leave, does the
network not persist?  I haven't experimented with this yet.



signature.asc
Description: OpenPGP digital signature
___
Sugar-devel mailing list
Sugar-devel@lists.sugarlabs.org
http://lists.sugarlabs.org/listinfo/sugar-devel


Re: [Sugar-devel] [Design] Ad-hoc networks - New Icons

2009-08-11 Thread Daniel Drake
2009/8/11 Benjamin M. Schwartz bmsch...@fas.harvard.edu:
 Hmm... so if other people join that network, and then I leave, does the
 network not persist?  I haven't experimented with this yet.

yes, it will persist
___
Sugar-devel mailing list
Sugar-devel@lists.sugarlabs.org
http://lists.sugarlabs.org/listinfo/sugar-devel


Re: [Sugar-devel] [Design] Ad-hoc networks - New Icons

2009-08-11 Thread Eben Eliason
On Tue, Aug 11, 2009 at 11:22 AM, Simon Schampijersi...@schampijer.de wrote:
 On 08/11/2009 03:49 PM, Eben Eliason wrote:

 On Tue, Aug 11, 2009 at 6:21 AM, Simon Schampijersi...@schampijer.de
  wrote:

 On 08/11/2009 11:50 AM, Daniel Drake wrote:

 2009/8/11 Simon Schampijersi...@schampijer.de:

 I think it would help, to have a new icon for the ad-hoc network to
 distinguish them. Could be a badged wireless network one? Or is the
 mesh
 icon appropriate? Or something completely new?

 I think new icons would be best, to distinguish from the mesh. I think
 we can expect mesh support again soon ;)

  From the user POV they are the same I guess. A local network, that does
 not
 need any infrastructure.

 Though, the mesh on the XO is handled automatically, the ad-hoc network
 requires user interaction to create it. I wonder if we ever will see a
 user
 using both (not at the same time) on the same machine. To think about the
 visual clash, at least.

 Perhaps we could use the mesh icon with a little XO badge, to indicate
 that it's functionally similar to the real mesh, but enabled by a
 specific XO. Thinking about this now, it might be the case that Tomeu
 had built this functionality as an extension of the wireless network
 device in the Frame; Should it be an extension of the mesh device
 instead, based on its perceived similarities to that feature more than
 an AP?

 Eben

 Yes, as of today, it is an extension of the wireless network frame device.
 http://1.bp.blogspot.com/_OoKpv4QinxI/SiFiDO2RsAI/ACU/go8n8S6rrE0/s1600-h/soas-create.png

 From the similarities, I agree, a badged mesh icon would work well to
 demonstrate that.

I suppose I see arguments in both directions. A badged AP icon would
also make sense. Ben's question about the persistence of the network
in the absence of the creator is also important to answer.

 Another question is the behavior: Gary and some others were wondering if we
 should fallback to an adhoc network automatically, if we are not connected
 to an AP.

This might bias us towards treating it more or less like the mesh. For
what it's worth, it seems like the ability for separate classes (for
instance) to create separate networks would be a benefit in terms of
network reliability.

Eben
___
Sugar-devel mailing list
Sugar-devel@lists.sugarlabs.org
http://lists.sugarlabs.org/listinfo/sugar-devel


Re: [Sugar-devel] [Design] Ad-hoc networks - New Icons

2009-08-11 Thread Benjamin M. Schwartz
Daniel Drake wrote:
 2009/8/11 Simon Schampijer si...@schampijer.de:
 I wonder if the ad-hoc network will scale up to that number of users :/
 Though I am not an expert in this area. Maybe Daniel has some more
 insights on this topic.
 
 Yes. Ad-hoc networks do not scale at all because they are range limited.
 
 They also act quite odd in terms of networks because you frequently
 have situations where C can see A and B, but A and B are unaware of
 each others presence. This will lead to funky situations where C could
 share an activity, A and B could join without problems, but any
 actions by A would not be seen by B and vice-versa.

This is a very important point, and has the potential to create a lot of
problems.  It also seems far from the conceptual model implied by the UI.
 If ad hoc networks are shown in the UI as being owned by a single user,
then it seems to me that what we want is not really an ad-hoc network at all.

The UI seems to be a much better description of the case in which single
user switches into Master mode and acts as an AP.  This user forwards
packets between other users to ensure mutual visibility as long as the
network owner is visible.

--Ben



signature.asc
Description: OpenPGP digital signature
___
Sugar-devel mailing list
Sugar-devel@lists.sugarlabs.org
http://lists.sugarlabs.org/listinfo/sugar-devel


Re: [Sugar-devel] [Design] Ad-hoc networks - New Icons

2009-08-11 Thread Peter Robinson
 Hmm... so if other people join that network, and then I leave, does the
 network not persist?  I haven't experimented with this yet.

 yes, it will persist

Really? My understanding of the ad-hoc network is that it basically
puts the user that creates it wifi card into ad-hoc AP mode and that
machine basically acts as a AP and link local IPs are used.

Peter
___
Sugar-devel mailing list
Sugar-devel@lists.sugarlabs.org
http://lists.sugarlabs.org/listinfo/sugar-devel


Re: [Sugar-devel] [Design] Ad-hoc networks - New Icons

2009-08-11 Thread Gary C Martin
On 11 Aug 2009, at 16:11, Daniel Drake wrote:

 2009/8/11 Simon Schampijer si...@schampijer.de:
 From the user POV they are the same I guess. A local network, that  
 does not
 need any infrastructure.

 I disagree. The mesh connections are automatic, and the presence of
 them does not indicate the presence of another computer like an ad-hoc
 network would do. Also, they do not have the principle of ownership
 that sugar places on ad-hoc networks.
 The behavioural properties of the networks (including the likelihood
 of communication) are different because there is no forwarding of
 frames. Also, neighbouring but sleeping laptops will not forward
 frames on an ad-hoc network.

H. Are you sure this is an accurate statement? I was under the  
impression that mesh forwarding support had been removed/disabled from  
OLPCs implementation a long time ago, since soon after the Mongolia  
deployment. Mesh was killing the wireless spectrum with all the  
attempted packet retransmissions. It is really only 'mesh' in name,  
all devices have to be in range of each other to collaborate.

Even with forwarding disabled it still didn't scale as well as hopped,  
so the next fallback plan was to recommend conventional infrastructure  
mode and APs for school size deployments.

Regards,
--Gary
___
Sugar-devel mailing list
Sugar-devel@lists.sugarlabs.org
http://lists.sugarlabs.org/listinfo/sugar-devel


Re: [Sugar-devel] [Design] Ad-hoc networks - New Icons

2009-08-11 Thread Daniel Drake
2009/8/11 Peter Robinson pbrobin...@gmail.com:
 Hmm... so if other people join that network, and then I leave, does the
 network not persist?  I haven't experimented with this yet.

 yes, it will persist

 Really? My understanding of the ad-hoc network is that it basically
 puts the user that creates it wifi card into ad-hoc AP mode and that
 machine basically acts as a AP and link local IPs are used.

Yes, really.
As an IBSS station you are responsible for attempting to send 10
beacons every second (including randomized backoff timer), but
cancelling each one if you hear another beacon beforehand. Therefore
in a nice RF environment, one person is the beacon sender sending 10
beacons every second (the station with the fastest clock) and the
others continually cancel their own beacon transmissions just before
they are about to transmit their own.

When the beaconing station goes away, in theory the person with the
2nd fastest running clock sends a beacon and then continues. the other
stations in the set hear that and keep synchronized.

Daniel
___
Sugar-devel mailing list
Sugar-devel@lists.sugarlabs.org
http://lists.sugarlabs.org/listinfo/sugar-devel


Re: [Sugar-devel] [Design] Ad-hoc networks - New Icons

2009-08-11 Thread Simon Schampijer
On 08/11/2009 05:42 PM, Daniel Drake wrote:
 2009/8/11 Simon Schampijersi...@schampijer.de:
 I wonder if the ad-hoc network will scale up to that number of users :/
 Though I am not an expert in this area. Maybe Daniel has some more
 insights on this topic.

 Yes. Ad-hoc networks do not scale at all because they are range limited.

 They also act quite odd in terms of networks because you frequently
 have situations where C can see A and B, but A and B are unaware of
 each others presence. This will lead to funky situations where C could
 share an activity, A and B could join without problems, but any
 actions by A would not be seen by B and vice-versa.

 The basic philosophy behind ad-hoc networks is shout unless someone
 else is shouting which also ends up causing various network splices
 and merges thanks to the poor quality of radios in the world, and
 interference.

 In short, ad-hoc networks are not reliable (especially when there are
 more than 2 participants) and are very quirky. It is a nice feature to
 have in sugar but it is not something that should be created
 automatically since in many cases they will just waste airtime for
 reliable networks and cause headaches for users.

Ok, agreed. I think the ad-hoc network is nice when for example, kid A 
meets kid B at home and they create a network they can communicate with.

Or another nice functionality is internet connection sharing. A (two 
network interfaces) is connected to the internet and creates an ad hoc 
network B can connect to. B will have internet access now as well.

 Automatic infrastructure-less networks are still very desirable of
 course, and the best solution that I know of for that would be
 open80211s which needs work done lower in the stack before it can be
 directly usable by sugar.

 and we're miles off topic on a thread about icons.. sigh
 Daniel

No, I think this discussion is needed to fully land this feature, and of 
course the icons.

Thanks for sharing your thoughts with us,
Simon

___
Sugar-devel mailing list
Sugar-devel@lists.sugarlabs.org
http://lists.sugarlabs.org/listinfo/sugar-devel


Re: [Sugar-devel] [Design] Ad-hoc networks - New Icons

2009-08-11 Thread Daniel Drake
2009/8/11 Gary C Martin g...@garycmartin.com:
 H. Are you sure this is an accurate statement? I was under the
 impression that mesh forwarding support had been removed/disabled from OLPCs
 implementation a long time ago, since soon after the Mongolia deployment.
 Mesh was killing the wireless spectrum with all the attempted packet
 retransmissions. It is really only 'mesh' in name, all devices have to be in
 range of each other to collaborate.

Yes, forwarding still happens.

And the mesh does scale quite well for sparse setups (its original
design). It also works quite well in dense setups (e.g. classrooms) in
that it allows reliable communication between about 15 nodes -- that's
about 13 more than we were able to do reliably with the other
infrastructure-free networking option (IBSS/ad hoc). of course,
classrooms of that size (that are additionally RF-space isolated from
other XOs) are not very common so we need other solutions there.

Daniel
___
Sugar-devel mailing list
Sugar-devel@lists.sugarlabs.org
http://lists.sugarlabs.org/listinfo/sugar-devel


Re: [Sugar-devel] [Design] Ad-hoc networks - New Icons

2009-08-11 Thread Gary C Martin
Hi Daniel,

On 11 Aug 2009, at 17:19, Daniel Drake wrote:

 2009/8/11 Gary C Martin g...@garycmartin.com:
 H. Are you sure this is an accurate statement? I was under the
 impression that mesh forwarding support had been removed/disabled  
 from OLPCs
 implementation a long time ago, since soon after the Mongolia  
 deployment.
 Mesh was killing the wireless spectrum with all the attempted packet
 retransmissions. It is really only 'mesh' in name, all devices have  
 to be in
 range of each other to collaborate.

 Yes, forwarding still happens.

Many thanks for that correction! I wish I'd seen some test data for  
these scenarios, with access to 3 XOs myself I'm very curious to go  
perform some tests now that you've confirmed it's not disabled  
(machine A, B C all on mesh, but with A and C out of direct wireless  
range, using B as the hop).

 And the mesh does scale quite well for sparse setups (its original
 design). It also works quite well in dense setups (e.g. classrooms) in
 that it allows reliable communication between about 15 nodes -- that's
 about 13 more than we were able to do reliably with the other
 infrastructure-free networking option (IBSS/ad hoc).

Ouch, well glad to hear testing was done on ad-hoc, that was before my  
time following OLPC.

 of course,
 classrooms of that size (that are additionally RF-space isolated from
 other XOs) are not very common so we need other solutions there.

So. From a design stand point, the tested, technical issues direct us  
to focus on making the creation of small ad-hoc groups (of just 2?)  
the sweet UI spot.

If you exclude the A can see B and C, but B cant see C case, does ad- 
hoc really fail so badly for ~5 kids in a room?

Regards,
--Gary

___
Sugar-devel mailing list
Sugar-devel@lists.sugarlabs.org
http://lists.sugarlabs.org/listinfo/sugar-devel