[h323plus] h323ep.cxx - Number translation problems.

[Jan Willamowius]

Hi Francisco,

I would agree that this kind of transformation should not be forced onto
every application.

My suggestion would be that you make a patch so applications can decide
if they want the transformation or not. And then we add the patch into
H323Plus instead of you keeping a permanent diff.

Simon, maybe you can weight in. I'm pretty sure you know applications
that rely on the current behavior. ;-)

Regards,
Jan

-- 
Jan Willamowius, jan at willamowius.de, http://www.gnugk.org/


Francisco Olarte wrote:
> A little background info: We use yate with h323plus to make calls,
> currently h323+ 1.25.0, yate 5.5.
> 
> In our system we use '#' extensively to delimit routing info ( as we
> use variable length prefixes, this makes them unique and we can
> distinguish between 10+347... and 103+47....).
> 
> But one day we used ## in a special prefix and the calls began to
> fail. We saw yate thar when dialing "0##666#88820334661" we ended up
> with the following ( every phne number redacted, cut last six numbers
> of phones for privacy ) pdu in the gatekeeper:
> 
> admissionRequest {
>         requestSeqNum = 32114
>         callType = pointToPoint <<null>>
>         endpointIdentifier =  12 characters {
> ....
>         }
>         destinationInfo = 1 entries {
>           [0]=url_ID "666%2388820334661 at 0"
>         }
> .....
>       }
> 
> which lead to a rejected call. We expected something on the line (
> from another PDU without double hash ):
> 
> admissionRequest {
> ...
>         destinationInfo = 1 entries {
>           [0]=dialedDigits "100#00543414"
>         }
> ...
> }
> 
> like:
> 
>         destinationInfo = 1 entries {
>           [0]=dialedDigits "0##666#88820334661"
>         }
> 
>  an url instead of "0#34666666666" to the gatekeeper. After further
> examination we found the following in h323+ 1.25.0:
> 
> [src]$ fgrep -R -C7 '"##"' h323plus/
> h323plus/src/h323ep.cxx-PBoolean H323EndPoint::ParsePartyName(const
> PString & _remoteParty,
> h323plus/src/h323ep.cxx-                                  PString & alias,
> h323plus/src/h323ep.cxx-
> H323TransportAddress & address)
> h323plus/src/h323ep.cxx-{
> h323plus/src/h323ep.cxx-  PString remoteParty = _remoteParty;
> h323plus/src/h323ep.cxx-
> h323plus/src/h323ep.cxx-  // Support [address]##[Alias] dialing
> h323plus/src/h323ep.cxx:  PINDEX hash = _remoteParty.Find("##");
> h323plus/src/h323ep.cxx-  if (hash != P_MAX_INDEX) {
> h323plus/src/h323ep.cxx-    remoteParty = "h323:" +
> _remoteParty.Mid(hash+2) + "@" + _remoteParty.Left(hash);
> h323plus/src/h323ep.cxx-    PTRACE(4, "H323\tConverted " <<
> _remoteParty << " to " << remoteParty);
> h323plus/src/h323ep.cxx-  }
> h323plus/src/h323ep.cxx-
> h323plus/src/h323ep.cxx-  //check if IPv6 address
> h323plus/src/h323ep.cxx-  PINDEX ipv6 = remoteParty.Find("::");
> 
> Well, it seems some function which should ( IMO ) be in the gatekeeper
> or the dialer as been grafted inside the generic endpoint code. Not
> too big a problem. We just developed a patch to zap the code patch and
> followed on. But then, we had a look ar h323plus 1.25.4 and found:
> 
>   PString remoteParty = _remoteParty;  // Support [address]##[Alias] dialing
>  PINDEX hash = _remoteParty.Find("##");
>  if (hash != P_MAX_INDEX) {
>    remoteParty = "h323:" + _remoteParty.Mid(hash+2) + "@" +
> _remoteParty.Left(hash);
>    PTRACE(4, "H323\tConverted " << _remoteParty << " to " << remoteParty);
>  }  PINDEX phash = _remoteParty.Find("#");
>  if (phash != P_MAX_INDEX) {
>     remoteParty.Replace("#","%");
>    PTRACE(4, "H323\tAdjusted " << remoteParty);
>  }  //check if IPv6 address
>  PINDEX ipv6 = remoteParty.Find("::");
> 
> Which is worst, now we can even use a single #. I wonder what is the
> rationale for doing this kind of transformations. We are developing
> another patch for this, but fear the future. For me it looks like this
> is trying to overcome the limitations of some device so the user can
> dial URLs ? IP ? from the phone keyboard, but on the way they are
> destroying our routing, and as they are buried deep inside the lib (
> where I personally think they shouldn't be ) they are tricky to detect
> and fix.
> 
> Shouldn't this kind of legal-number-destroying functionality be
> relegated to an overridable method, or just delegated to the client
> code? Is this going to go on ( I mean, more exotic translations on
> otherwise legal dialedDigits ) ? And I'm not the only one using '#' in
> the digits, some of our providers do it too.
> 
> Regards.
>     Francisco Olarte.
>