> I'm pretty sure a DoS can be considered as an attack without even caring 
about any other hint. So if a hub (or an hub-impersonator) DoS a 
subscriber, the subscriber should take adequate measures wether there is 
(or not a verify token). 

OK, I'll concede that; you're right, a DoS has other characteristics that 
allow it to be identified.  Still, it seems useful in the subscribe/verify 
handshake between the subscriber and the hub, that the subscriber can 
verify the identity of the hub when processing the verification request.  

>But if the hub uses HTTPs to communicate with the subscriber how can 
anyone "snif" the callback url? since the callback url was only 
communicated over HTTPs to the hub, and then back to the subscriber, there 
is no way anyone can snif anything :) 

OK, here is where I have to apologize and admit to a mistaken understanding 
on my part.  I was under the impression that the URL was exposed during an 
HTTPS communication, and only the request content was encrypted.  However, 
since you obviously believed the opposite to be true, I fell back to the 
old poker adage "if you've been in a poker game for a while and you don't 
know who the sucker is, get up, it's you".  And so I looked it up and guess 
what, it IS me :-)

So, mea maxima culpa and all that.  In my defense, I think I said several 
messages back that I'm not a security expert, and now I've proven it.

And so, I admit that if the subscriber uses HTTPS, the callback URL will 
remain secret.  This may put a bit of extra burden on some subscribers (in 
terms of getting and managing a cert if they're not already doing so), but 
it's a solution.  I hadn't personally been using HTTPS for my callback, but 
I certainly could.

Thanks again for the exchange!

-Andy

Reply via email to