> On Feb 23, 2017, at 7:59 PM, Aaron Falk <[email protected]> wrote:
> 
> On 23 Feb 2017, at 13:33, Michael Welzl wrote:
>>> On Feb 23, 2017, at 7:05 PM, Tommy Pauly <[email protected]> wrote:
>>>> On Feb 23, 2017, at 12:55 AM, Michael Welzl <[email protected]> wrote:
>>>>> On Feb 22, 2017, at 3:54 PM, Aaron Falk <[email protected]> wrote:
>>>>> 
>>>>> Trimming for focus…
>>>>> 
>>>>>>> 
>>>>>>> What are the key questions TAPS should be focusing on at this meeting?
>>>>>>> 
>>>>>> I can think of these 3 major questions that appeared to us when we 
>>>>>> worked on our drafts:
>>>>>> 
>>>>>> - how do we handle multi-streaming?
>>>>>> - how do we handle messaging?
>>>>>> - what about message sizes?
>>>>> 
>>>>> These seem like good discussion topics.  Who would like to lead a short 
>>>>> discussion on each?
>>>> 
>>>> I’d volunteer for all 3. Priority order:
>>>> 1. messaging - I’d really like to lead the discussion on this one
>>>> 2. multi-streaming
>>> 
>>> I'd like to volunteer to lead the discussion about how to handle multi 
>>> streaming protocols. We've been working on that and giving it a lot of 
>>> thought for our protocol stack.
>> 
>> Speaking as someone who volunteered for the same: I’d be absolutely fine 
>> with that.
>> 
> 
> So, seems like we still could use someone to lead a discussion about message 
> sizes.  I confess I don’t recall what the concern is.  Michael, can you 
> expand a little?

UDP applications (at least) sometimes need to know the PMTU - e.g. that was on 
the list of things a latency-critical application needs, presented by Colin at 
the last meeting.
The PMTU is a per-path property. Should a TAPS system be fully aware of a 
physical path? If so, consider: the PMTU is including all headers - so the 
message size that fits in will be different for v4 and v6.
For the real message size that fits in, subtract the UDP header too. Then we’re 
dependent on layers 3 AND 4 … not how a TAPS system should be!  Yet, for apps 
that want to do PMTUD (common above UDP), an app *must* be able to send a 
packet exceeding the MTU… how should all this be handled / exposed in a TAPS 
system?

Cheers,
Michael

_______________________________________________
Taps mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/taps

Reply via email to