Re: [6lowapp] [6lowpan] hardware trends, new vs. existing protocols [Re: 4861 usage in LLNs]

[Vlad Trifa <[e-mail]>]

I totally second you Robert!

I think indeed, we shouldn't try to accomodate everything from enocean  
light switches to servers, but rather enabling interoperability  
between a wide class of devices with similar (limited) resources,  
basically the same class of devices where 6lowpan could be used, and  
basta.

Indeed, our focus should be really making the web of things a reality,  
and make it scale as well, rather than take into account too much  
bandwidth constraints and have them as central "design problem to  
solve", as the actual problem we're trying to solve is different, and  
bandwidth is a not-so-critical constraint (if we need very high data  
rate and quasi real time constraints, nobody prevents us by using  
gateways to connect legacy networks with other pure CoRE ones), as  
long as CoRE can "semantically" accommodate different QoS with high  
data rates transmitted through protocols that we're designed for that  
(http). If the proxying or device-level core-enablement is  
transparent, then it doesn't matter because interactions will be  
CoREful (I like that term :)  anyway.

It's not because a protocol was not made for something that we  
should't use it, especially if it works fine for the scenarios we  
consider. What's really missing is a concrete evaluation of the  
scalability of HTTP-based scenarios for devices to show that it works  
(or doesn't). Unless we have that, I don't see why we start with the  
assumption that http-based solutions would be inherently bad and won't  
scale because of the verbosity, and we need to get a grasp of their  
actual limitations in that area.

Vlad




On Nov 11, 2009, at 1:25 AM, Robert Cragie wrote:
> Regarding resource: There will be classes of devices which currently  
> will not support IP packets as the resource constraints are too  
> high, e.g. self-powered switches where the power budget is so  
> constrained you can barely get a squeak out of them. Do we want to  
> accommodate these devices in CoRE etc.? Probably not, and so it  
> makes sense to develop the network which suits those devices and  
> proxy them on a gateway. On the other hand, I think we have  
> established that we don't want to persistently continue to develop a  
> plethora of coexisting but non-interoperating networks all connected  
> to the internet through a hodge-podge of application gateways for  
> the sake of it. This is the reality today and it could be argued  
> that this has stifled the development of the highly-connected  
> "Internet of Things" we have all been dreaming about; it is a  
> solution but a clumsy one and one which doesn't scale well.
>
> Regarding bandwidth: I remember back in the early 90's being able to  
> look at basic websites over a V32.bis (14400 bps max.) modem; whilst  
> the experience could be frustratingly slow, it was usable. And that  
> was an online, interactive experience. So whilst we need to consider  
> the arguments about bandwidth, I think that realistic traffic  
> scenarios need to be carefully looked at before ruling some  
> solutions out. I am sure there is already a plethora of experience  
> from many of the contributors to this group who already work in this  
> area.
>
> So we come down to the limited ROM/RAM devices which sit uneasily  
> between the clearly capable devices currently available and the very  
> constrained and highly specific devices for certain types of  
> network. Which way do we want to push them? On the basis that these  
> are the devices which will become obsolete first, I would push them  
> towards the constrained side and say they are all proxied. In which  
> case, CoRE/ROLL/6LoWPAN becomes becomes focused on the more clearly  
> capable devices and therefore, as said at the beginning of this e- 
> mail, based on protocols taken verbatim where possible and adapted  
> where necessary.
>
> Robert
>
> Kris Pister wrote:
>>
>> Richard -
>> I think that today's things are being designed with wonderful chips  
>> like your Ember EM351 and EM357
>> which have 128kB and 192kB of flash and lots of RAM; like the  
>> Jennic JN5148, the Freescale MC13224, the Dust DN2510.
>> They can run IP, they will run IP, and in many cases they do run  
>> IP.  We all agree on that, and we're all excited about that. The  
>> debate centers on how many new protocols we need to invent, vs. how  
>> many we can adopt or adapt, with the existing hardware, and with an  
>> eye toward where technology trends are taking us.  My concern, like  
>> yours, is over the rate of adoption.  If the fastest path to broad  
>> adoption is to create new protocols for routing, ND, transport, and  
>> applications, then by all means let's do that.  I'm concerned,  
>> however, that this has not been a uniformly successful approach for  
>> wireless sensor networks in the past. :)
>> Many of us believe that we will see the fastest adoption by  
>> minimizing the number of new protocols.  We might be wrong, and  
>> that's the debate.
>>
>> ksjp
>>
>> Richard Kelsey wrote:
>>>    Date: Mon, 09 Nov 2009 22:12:03 -0800
>>>    From: Kris Pister <[e-mail]>
>>>
>>>     > Abandoning the installed base just goes to reinforce the idea
>>>     > that IP isn't an appropriate technology for things.
>>>
>>>    Michael - I think that we have the same goal, but I disagree  
>>> with that    statement.  I think that re-writing every protocol  
>>> from discovery    through transport to applications, from scratch,  
>>> is what reinforces the    idea that IP isn't an appropriate  
>>> technology for things.
>>>
>>>    I realize that there are pressures from an installed base, but  
>>> at this    point it's a tiny fraction of the overall potential.   
>>> If we let the 1%    installed base dictate the path for the next  
>>> 99%, we should do our best    to ensure that it's the right path.
>>>
>>> Taking these two paragraphs together, you seem to be saying
>>> that IP is an appropriate technology for tomorrow's things,
>>> but not necessarily for today's.  While the hardware will
>>> obviously improve over time, we still need to pick some
>>> target platform.  The current 6lowpan charter gives 32K of
>>> flash as an example and mentions 802.15.4 repeatedly.  Are
>>> you suggesting that we recharter?
>>> The increasing capabilities of the hardware does give us the
>>> reassuring prospect that the longer we take the solve the
>>> problems the easier it will be to so.
>>>
>>>                                 -Richard Kelsey
>>>
>> _______________________________________________
>> 6lowapp mailing list
>> [e-mail]
>> https://www.ietf.org/mailman/listinfo/6lowapp
>>
> _______________________________________________
> 6lowapp mailing list
> [e-mail]
> https://www.ietf.org/mailman/listinfo/6lowapp