Re: [edu-team] So....we can do a wireless tutorial, with Donald Eastlake presenting

[Brian E Carpenter <[e-mail]>]

On 10/10/2013 09:30, Radia Perlman wrote:
> Brian...How *does* the use of ARQ in 3GPP and WiFi affect whether or not we
> need a UDP checksum?
> 
> ...since you mentioned it...I don't know what you'd consider a good answer
> to that...
> 
> do you mean why do we have checksums at multiple layers?

As I understand it, both WiFi and 3GPP go to considerable lengths
to retransmit packets that arrive with errors, and in fact that
leads to great variability in packet latency on links with high error
rates - not something we see these days on wired links. So all the
folk knowledge that led to dropping the header checksum in IPv6 but
mandating the UDP checksum is obsolete. If the least reliable link
layer is doing ARQ, why bother with layer 3 and 4 checksums?

But on the other hand, the Coded TCP work shows that in some
circumstances, e2e performance is significantly better if you disable
ARQ in wireless layer 2 and do the error detection only in layer 4.

Also there might be applications that would prefer lower latency and
some errored packets to variable latency and error-corrected packets.

It was only an example where one part of our technology affects
another, which is where I think EDU should focus.

    Brian

> 
> Radia
> 
> 
> On Wed, Oct 9, 2013 at 12:45 PM, Brian E Carpenter <
> [e-mail]> wrote:
> 
>> Radia,
>>
>> I am concerned that we've had a tendency to run tutorials whose
>> teachers haven't really considered the guidelines that I just
>> drafted. That's really all - and since we'd never written down
>> the guidelines, it isn't even surprising. So if Donald explains
>> what transport protocol designers, application protocol designers,
>> compression algorithm designers, etc. need to understand about
>> wireless, that would be great. How does the use of ARQ in 3GPP
>> and WiFi affect whether or not we need a UDP checksum? Etc.
>>
>> Donald's very likely to do this without being told, but I'm not
>> sure that applies to all speakers.
>>
>> Regards
>>    Brian
>>
>> On 09/10/2013 15:15, Radia Perlman wrote:
>>> Personally, I think helping people catch up on various technology will
>>> definitely "help IETF work better".  Even making people who supposedly
>> know
>>> an area rethink the technology in order to summarize the concepts, the
>>> controversies, etc., will also "help IETF work better". Having people
>>> outside an area be able to understand other areas, both in order to lend
>>> their expertise, and in order to understand the implications of other
>>> technology, will also "help IETF work better".
>>>
>>> Even the most technically expert IETF person tends to only know about a
>>> subset of the technical work in IETF. Sure, they could read the mailing
>>> list archives and all the internet drafts for other areas...in
>> theory...but
>>> having them attend a 2 hour tutorial instead, will "help the IETF work
>>> better".
>>>
>>> So I am totally befuddled by what you think the phrase "help IETF work
>>> better" means.
>>>
>>> If you mean only procedural tutorials (e.g., how to use the tools, what
>> the
>>> rules of IETF are, etc.), I think those are useful too.  But certainly,
>>> facilitating understanding technology, fostering cross-fertilization,
>>> enabling newcomers to have something that they can sit in on, on day one,
>>> and understand, definitely also "helps IETF work better".
>>>
>>>
>>> Radia
>>>
>>>
>>> On Tue, Oct 8, 2013 at 5:45 PM, Brian E Carpenter <
>>> [e-mail]> wrote:
>>>
>>>> Radia,
>>>>
>>>> I think you ducked my question. I have taught some of this stuff in
>>>> the past, and I recently heard a fascinating talk by Muriel Médard
>>>> about how Coded TCP can deal with lossy wireless links. It's very
>>>> interesting. But my question for any EDU-sponsored tutorial, whether
>>>> on a technical or a procedural topic, is: how will this make the
>>>> IETF work better?
>>>>
>>>> (In this case there's a corollary: did the outcome from the PILC
>>>> WG a few years ago produce any benefits for IETF protocols?)
>>>>
>>>> To be clear, given the short notice in this case, it seems like
>>>> a good way to fill the gap. But I think EDU needs more of a strategic
>>>> plan with tight focus on making the IETF work better. That affects
>>>> the topics we pick and the direction we give to the speakers.
>>>>
>>>> Positive thought: maybe we need a standard writeup that we send to
>>>> all potential speakers. Having said that, I suppose I'd better
>>>> draft it...
>>>>
>>>>     Brian
>>>>
>>>> On 06/10/2013 08:55, Radia Perlman wrote:
>>>>> Wireless links have somewhat different properties than wired links, and
>>>>> there are fascinating implications.  For instance, TCP is impacted.
>>  It's
>>>>> important to understand the properties of the underlying technology.
>>>>>
>>>>> I was thinking of suggesting a title like
>>>>> "Wireless links; Properties, Challenges, Solutions, and Implications"
>>>>>
>>>>>
>>>>> On Sat, Oct 5, 2013 at 12:07 PM, Brian E Carpenter <
>>>>> [e-mail]> wrote:
>>>>>
>>>>>> I hate to be ornery, but which current IETF topics will benefit
>>>>>> (in the sense that WGs will be able to do a better job)?
>>>>>>
>>>>>> Regards
>>>>>>    Brian
>>>>>>
>>>>>> On 06/10/2013 06:07, Radia Perlman wrote:
>>>>>>> He gave a wireless tutorial somewhere that I watched, perhaps 10
>> years
>>>>>> ago,
>>>>>>> and it was excellent.  He said that some people from 802.11 and
>>>>>>> Bluetooth/Zigbee have been helping him update it, so he is all set
>> for
>>>>>> 2013
>>>>>>> Vancouver, and can do either slot (1 or 3 PM).  I think we should go
>>>>>> ahead
>>>>>>> and schedule it.
>>>>>>>
>>>>>>> Radia
>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>>
>>>> ------------------------------------------------------------------------
>>>>>>> _______________________________________________
>>>>>>> edu-team mailing list
>>>>>>> [e-mail]
>>>>>>> https://www.ietf.org/mailman/listinfo/edu-team
>>
>