RE: EXP2ST: TSI length limitations with A flag - LCT/ALC

You make some good points Jani, and thus given this I think Rod's suggestion
is the best way to go unless anybody thinks of something better.

-----Original Message-----
From: Jani Peltotalo [mailto:[email protected]] 
Sent: Monday, August 08, 2005 11:13 PM
To: 'Michael Luby'; [email protected]; [email protected]
Subject: RE: [Rmt] EXP2ST: TSI length limitations with A flag - LCT/ALC

Hi all,

>From implementation point of view it is important to see the header field
lengths from the header flags (e.g. S, O, H). In our implementation receiver
parses TSI and TOI values as informed with S, O and H flags, so if H flag is
set to one (and O is zero) then our receivers tries to parse TOI value also.
Then if there is no TOI field present, receiver needs extra information to
check if there is a header or packet error. The 'A flag' is not enough
because you can use that in non-empty packets also. And I think that we
should avoid extra complexity.      

Another side effect is that current receiver does not interoperate with
suggested idea.

Cheers, Jani

> -----Original Message-----
> From: [email protected] [mailto:[email protected]] On Behalf Of
> Michael Luby
> Sent: 2. elokuuta 2005 2:42
> To: [email protected]; [email protected]
> Subject: RE: [Rmt] EXP2ST: TSI length limitations with A flag - LCT/ALC
> 
> Another option is to just relax the requirement that the TOI + TSI must be
> 32-bit aligned only in the case when the remainder of the packet is
> non-empty (carries data).  The alignment is really for accessing the data
> in
> the payload of the packet, but if there is no data then the alignment does
> not seem important.
> 
> -----Original Message-----
> From: [email protected] [mailto:[email protected]] On Behalf Of
> [email protected]
> Sent: Sunday, July 31, 2005 10:04 AM
> To: [email protected]
> Subject: [Rmt] EXP2ST: TSI length limitations with A flag - LCT/ALC
> 
> TOI may be omitted for A flag packets with FLUTE (and also for a single
> object transfer with ALC). LCT says TOI + TSI filed must be 32 bit
> aligned.
> TSI may be 16, 32 or 48 bits length. Thus, the TOI field may not be
> omitted
> for objects in sessions with TSI lengths of 16 or 48 bits. (for 16 bits,
> the
> TSI can be increased to 32 bits just for A flag packets without TOI by
> switching S, O, H flags - but this is not desirable and doesn't work for
> 48bit TSI).
> 
> Should we add a note to LCT (and/or ALC) that "Omitting the TOI field is
> only feasible for packets with 32 bit TSI field length."?
> 
> BTW, an implementation option is to set TOI = 0 for A flag packets where
> TSI
> is 16 or 48 bits length
> 
> Cheers, Rod.
> 
> _______________________________________________
> Rmt mailing list
> [email protected]
> https://www1.ietf.org/mailman/listinfo/rmt
> 
> 
> 
> _______________________________________________
> Rmt mailing list
> [email protected]
> https://www1.ietf.org/mailman/listinfo/rmt