Re: [Spud] [Int-area] New Version Notification for draft-welzl-icmp-text-middleboxes-00.txt

[Michael Welzl <[e-mail]>]

[ + spud list ]


> On 2. jul. 2015, at 22.07, Black, David <[e-mail]> wrote:
> 
>>>>> cases where the latter doesn't happen.
>>>> 
>>>> Got it now, sorry for the confusion.
>>>> So I don't really care much about the middlebox behind a NAT on the other
>> side - it's *my* side I care about. The usage examples in the draft refer to
>> talking people in my institution or at the airport where I'm sitting. I don't
>> expect to be (or care about being) able to talk to a middlebox far, far
>> away... chances are, the message is dropped long before anyway.
>>> 
>>> It's a limitation; it might be useful to explain the limitation and why
>>> the remainder is still useful.
>> 
>> ok
> 
> And specify boundaries at which these text messages SHOULD be dropped because
> they are likely to be irrelevant beyond that point.  For the airport or
> university example, this may be where the airport or university network hands
> off to the ISP that provides connectivity to the Internet.

ok


> I also concur with Joel's and Diego's observation that an operator with strongly
> structured procedures (e.g., ticketing system) is going to have some difficulty
> dealing with this sort of problem report, especially as it's unauthenticated -
> the fact that one can get an ICMP packet into the operator's network does not
> automatically entitle one to draw upon the operator's support resources.

This was also said by Diego Lopez - such operators may just not be a suitable target for this type of message (rather, smaller ones, then). But "entitle to draw upon the resources" is a bit exaggerated when we're talking about sending an ascii message. Nobody's obliged to listen to it, after all.

Anyway, I have to say I'm getting some very good feedback here - in particular, I like Dan Wing's observation (on the spud list) that applications could auto-generate this (e.g. "Tried QUIC but failed UDP, falling back to TCP"); he suggested a machine-parsable format but now I wonder how to do this?

All of a sudden, this seems to become more reasonable and make more sense than I ever expected   :-)

Thanks!
Michael