[Insipid] session-id ABNF (Was: Backward compability between draft-jones and draft-kaplan)

[Hadriel Kaplan <[e-mail]>]

When we talked about this in Vancouver I was thinking there wouldn't be any problem going to/from UUID syntax, because going to/from UUIDs in H.323 was talked about a long, long time ago for session-id.  But I forgot about the bug in the draft-kaplan ABNF...

Since version 01 of draft-kaplan-sip-session-id in 2008, which later became draft-kaplan-dispatch-session-id, the rules for generating the session-id always made it lower-case *hex* characters.  I.e., the g-z range wasn't actually possible.  The ABNF for it didn't get fixed until draft-kaplan-dispatch-session-id-01 in April 2010, but that also dropped it to 16 characters of hex because the algorithm changed, and then 3 months later reverted back to 32 characters and the correct ABNF.

So that begs the question of *which* version of draft-kaplan we want to be backwards compatible with.  Since draft-kaplan-dispatch-session-id-01 only existed for 3 months, I think we can ignore that one.  And since everything else only used hex characters, except the initial 00-draft which itself only existed for 4 months, I think we can ignore that one too.

Therefore, as far as I can tell we shouldn't have a problem converting draft-kaplan session-ids to/from UUIDs for H.323/whatever.

-hadriel


On Aug 14, 2012, at 4:27 PM, Christer Holmberg wrote:

> Hi,
> 
> In Vancouver, I expressed a wish to make the INSIPID mechanism backward compatible with the draft-kaplan-session-id one, the reason being that it has been around for a relatively long time, and has been implemented and deployed.
> 
> I compared the suggested draft-jones solution with the draft-kaplan solution, and I think there are ways to achieve backward compability.
> 
> 
> --------------
> 
> 
> FIRST, looking purely at the ABNF syntax:
> 
> The jones ABNF is:
> 
> 
> Session-ID-UUID = "Session-ID-UUID" HCOLON UUID
> 
> 
> 
> The kaplan ABNF is:
> 
> 
> 
> Session-ID           =  "Session-ID" HCOLON sess-id
> 
>                           *( SEMI generic-param )
> 
> sess-id              =  32(DIGIT / %x61-7A)  ; 32 chars of [0-9a-z]
> 
> 
> 
> 
> 
> Now, assuming 32 chars is enough to carry ONE UUID, jones applications could put the SECOND UUID in a header field parameter (generic-param), rather than combining it with the received UUID.
> 
> 
> 
> Of course, the first UUDI (carried in sess-id) can only contains characters 0-9 and a-z.
> 
> 
> 
> A UUID can contain characters 0-9 and a-f and A-F.
> 
> 
> 
> So, a jones entity would have to be prepared to receive g-z.
> 
> 
> 
> And, as kaplan entity is not prepared to receive capital A-F, a jones entity would have to only use small letters. But, that should not be a problem, afaik, as they have the same meaning as far as UUID is concerned.
> 
> 
> 
> --------------
> 
> 
> 
> SECOND, looking at it would work functionwise:
> 
> 
> 
> There are two main alternatives:
> 
> 
> 
> 
> 
> 1. Second UUID is added even when one entity is kaplan compliant
> 
> 2. Second UUID is not added when one entity is kaplan compliant
> 
> 
> 
> 
> 
> Alternative 1:
> 
> 
> 
> When a kaplan entity inserts the first UUID (I use that as a general term, eventhough a kaplan entity does not use UUIDs), one must assume that it is prepared to receive a response from a jones entity which has added the second UUID into the header field parameter.
> 
> 
> 
> When a jones entity inserts the first UUID, one must assume that it is prepared to receive a response from a kaplan entity where the second UUID has *not* been added.
> 
> 
> 
> 
> 
> Alternative 2:
> 
> 
> 
> When a kaplan entity inserts the first UUID, the jones entity detects (using some identify-a-kaplan-entity mechanism) that it comes from a kaplan entity, so it does not insert a seoncd UUID in the first place.
> 
> 
> 
> As in alternative 1, when a jones entity inserts the first UUID, it has to be prepared to receive a response where the second UUID has not been added (the identify-a-kaplan-entity mechanism might also inform the jones entity that the response comes from a kaplan entity.
> 
> 
> 
> 
> 
> --------------
> 
> 
> 
> Now, there will of course be cases, where two UUIDs are required, that won't work. But, kaplan entities would be able to use the session-id for whatever they've used it so far.
> 
> 
> 
> 
> 
> Regards,
> 
> 
> 
> Christer
> 
> 
> 
> 
> 
> 
> 
> 
> _______________________________________________
> insipid mailing list
> [e-mail]
> https://www.ietf.org/mailman/listinfo/insipid