[ih] SIP and ENUM

[Karl Auerbach]

There are two topics in this thread:

- SIP peer-to-peer

- Enum

With regard to the former: One of the products I build is a tool that 
"does bad things" to traffic.  (Things like drops, jitter, packet 
resequencing, etc - in other words, a descendant of Jon's idea of a 
"flakeway" for use in the old TCP bakeoffs.)  For testing I often need a 
convenient traffic flow that is easy to evaluate for the impact of these 
"bad things".  I've found peer-to-peer SIP phones quite useful for this 
as it is often quite easy to hear the effects of those "bad things" 
(most SIP phones don't like the voice packets being jittered, or worse, 
re-ordered.  The audio artifacts are pretty obvious.)

(BTW, once I did a demo at the iLabs at an Interop show in which I 
introduced new words into a SIP call.  It wasn't hard.  I was 
considering going another step and see if I could detect occurrences of 
the word "not" and elide them from the conversation.  It seemed feasible 
but more work than I was then willing to put into the project.  The 
lesson is to make sure that conversations are encrypted.)

As Patrik F. suggests, some SIP phone software wants to use a SIP server 
in the middle.  But not all.  For instance, I use Grandstream phones for 
testing because they are happy to operate without a SIP server in the 
middle (or DHCP or much else.)

As for Enum - that has always scared me.  I am made nervous by the 
notion that DNS clients would evaluate j-random regular expressions 
found in NAPTR records.   That strikes me not much different than the 
security problems we have had with web browsers evaluating random 
Javascript fed to them by websites.

         --karl--