Paul,
I put together a spreadsheet to evaluate all possible combinations of fixes and versions. It's pretty small (15K), so I attached it.
You are right that there will be some v2 entities that expect to see the old aliasesInconsistent choice, but they will now only receive them from other v2 entities (compatible ones choices) or v3 entities that have not applied what I propose to be added to IGv3 (incompatible choices). The latter is simply broken, and there is nothing we can do to fix it, no matter which approach we take. A v2 entity will be no worse than a v1 entity in that it will receive a reject reason that it doesn't understand because it was added after v2, e.g., a new aliasesInconsistent. In this case, it should treat the reason as undefinedReason. This is the way unrecognized reasons and choices in general have always been treated.
Regarding v3, if we swap the choices in IGv3 and H.225.0v4 as you propose, "pre-fix" v3 entities like our EPs will definitely encounter decode errors--they will always be at risk. The same goes for v3 GKs. If a v3 GK transmits either of these pre-fix choices, the receiving "fixed" EP will encounter a decode error. On the other hand, the worse thing that can happen using my proposal is that an entity would treat these pre-fix choices as undefined reasons--a much better outcome.
We should also recommend that v2 GKs no longer send the old aliasesInconsistent reason. They should instead send undefinedReason. It is unfortunate that they would lose the ability to convey this particular info to the EP, but the alternative is risking a decode error and dropping the call.
I realize that my proposed fix of deprecating the two extant reasons is less elegant than simply switching them around in order to agree with v2. However, as you can see from the spreadsheet, the benefit of interoperability far outweighs this.
Paul Long Smith Micro Software, Inc.
-----Original Message----- From: Paul E. Jones [mailto:paulej@packetizer.com] Sent: Thursday, October 05, 2000 11:15 PM To: Paul Long; Mailing list for parties associated with ITU-T Study Group 16 Cc: h323implementors@imtc.org; h323implementors@pulver.com Subject: Re: Error in H.323v3 ASN.1
Paul,
The problem with this simpler approach to deprecating the fields is that there will still be v2 entities that expect to see the aliasesInconsistent field and there may be v3 entities out there that assume a different field is there and things will be broken between those endpoints.
Perhaps it might be less of an impact, though...
Paul
----- Original Message ----- From: "Paul Long" Plong@SMITHMICRO.COM To: "Mailing list for parties associated with ITU-T Study Group 16" ITU-SG16@mailbag.cps.intel.com Cc: h323implementors@imtc.org; h323implementors@pulver.com Sent: Thursday, October 05, 2000 2:08 PM Subject: RE: Error in H.323v3 ASN.1
Paul,
I just checked, and we are shipping v3 EPs. I agree that there are
probably
many more v2 entities than v3 entities, and therefore if we have to
break
one, we should break the one which will have the least impact on the
market.
To minimize the impact this will have on interoperability, however,
let's
don't simply move the reasons around. Instead, lets deprecate the
current
routeCallToSCN and aliasesInconsistent reasons using the following
syntaxes
in IGv3 and H.225.0v4.
Proposed IGv3 syntax
AdmissionRejectReason ::= CHOICE { calledPartyNotRegistered NULL, -- cannot translate address invalidPermission NULL, -- permission has expired requestDenied NULL, -- no bandwidth available undefinedReason NULL, callerNotRegistered NULL, routeCallToGatekeeper NULL, invalidEndpointIdentifier NULL, resourceUnavailable NULL, ..., securityDenial NULL, qosControlNotSupported NULL, incompleteAddress NULL,
-- changes start here: deprecatedAliasesInconsistent NULL, -- do not use deprecatedRouteCallToSCN SEQUENCE OF PartyNumber, -- do not use callCapacityExceeded NULL, -- destination has exceeded call capacity aliasesInconsistent NULL, -- multiple aliases in request identify
distinct
people routeCallToSCN SEQUENCE OF PartyNumber }
Proposed H.225.0v4 syntax
AdmissionRejectReason ::= CHOICE { calledPartyNotRegistered NULL, -- cannot translate address invalidPermission NULL, -- permission has expired requestDenied NULL, -- no bandwidth available undefinedReason NULL, callerNotRegistered NULL, routeCallToGatekeeper NULL, invalidEndpointIdentifier NULL, resourceUnavailable NULL, ..., securityDenial NULL, qosControlNotSupported NULL, incompleteAddress NULL,
-- changes start here: deprecatedAliasesInconsistent NULL, -- do not use deprecatedRouteCallToSCN SEQUENCE OF PartyNumber, -- do not use exceedsCallCapacity NULL, -- destination does not have the capacity
for
this call aliasesInconsistent NULL, -- multiple aliases in request identify
distinct
people routeCallToSCN SEQUENCE OF PartyNumber, collectDestination NULL, collectPIN NULL }
I assume we can change both because neither are Decided and these are
true
irreconcilable errors in the documents, correct? Here are the
specifics: We
rename the old reasons to "deprecate*", add the new ones after callCapacityExceeded/exceedsCallCapacity, and then in the text of IGv3
and
H.225.0v4 say that the old ones shall not be transmitted and if
received
shall be treated like undefinedReason. Otherwise, a receiver would
have to
parse the AdmissionReject message differently depending on the version
of
the transmitter. We, Smith Micro, can actually do that with our ASN.1
parser
but I'm guessing that most implementations cannot. Maybe we should
even
change the type of deprecatedRouteCallToSCN to NULL so that the ASN.1
PER
parser would simply skip over the contents via the encapsulating open
type.
I know that our parser would do this. Would others?
v1 implementations won't have a problem, because they would ignore
this
reason, anyway, but we could encourage v2 implementations to treat aliasesInconsistent like undefinedReason since they won't know if it
is
aliasesInconsistent from a v2 EP or some other extended choice from a
v3+
EP.
Paul Long Smith Micro Software, Inc.