[stir] Can canonical phone numbers survive SBCs and other middle boxes?

["Rosen, Brian" <[e-mail]>]

The fundamental idea that we have here is that you sign a set of information with a credential, and you pass the signature and a pointer to the credential in SIP signaling.

We have some disagreements about what is signed.

Henning has proposed that we canonicalize the From and To phone numbers, we include a timestamp and some form of call-id, possibly the INSIPID id.

There are assertions that you can't use From/To, because middle boxes change them.  Some have suggested using P-A-I and other headers instead.

Hadriel wrote a draft about why From and To get modified:
http://tools.ietf.org/id/draft-kaplan-sip-uris-change-00.txt

After reading it, I am unclear if a canonicalized e.164 would make it through.  The reasons given seem to indicate they would.  They change domains, they change prefixes, but they don't seem to change the actual telephone number.

Can we come up with examples where a canonicalized e.164 would NOT pass end to end?

Brian