SCTP Discussions

Text archives Help


Re: [SCTP-Discussion] About the SCTP checksum


Chronological Thread 
  • From: Luca Papotti <Luca.Papotti AT italtel.it>
  • To: discussion AT sctp.de
  • Subject: Re: [SCTP-Discussion] About the SCTP checksum
  • Date: Tue Jul 10 11:45:02 2001
  • List-id: SCTP Discussions <discussion.sctp.de>

Hi all,
i agree with the idea of an ULP digest, but this do not resolves the
problem (if it is a real problem) of unnecessary chunk(s) retransmission
at trasport level.
A potential alternative solution is an Adler-32 checksum covering all
control informations including

1. SCTP header
2. all chunk headers
3. All control chunks (SACK ...)

and a distinct checksum 4 every DATA CHUNK.

This sol. has the advantage to remove the probability of unnecessary
chunks retransmission, at the same computational cost of the overall chksm.
Obviously with the overhead of a checksum field (4 byte) in every DATA header.
What do You think ??



P.S. I'm not suggesting to modify the RFC 2960, it's only 4 discussion purpos
;)


Tuexen Michael ha scritto:

> Dear all,
>
> the SCTP checksum provides packet integrity, not DATA chunk integrity.
> Therefore the SCTP header is included in the checksum calculation.
> Something you won't have with only checksumming DATA chunks. And what
> about control chunks?
>
> So I think you need a fast checksum (or CRC) for the whole packet. If this
> is not enough for user data, the ULP can use a digest to provide higher
> integrity. This has not to be recalculated in case of retransmission.
> See for example
> http://www.ietf.org/internet-drafts/draft-tuexen-tsvwg-tls-over-sctp-00.txt
> Best regards
> Michael
>
> Michael Tuexen Tel.: +49 89 722 47210
> Siemens AG Fax: +49 89 722 48212
> ICN WN CS SE 5 E-mail:
> Michael.Tuexen AT icn.siemens.de
>
> > -----Original Message-----
> > From: Luca Papotti
> > [SMTP:Luca.Papotti AT italtel.it]
> > Sent: Monday, July 09, 2001 5:50 PM
> > To:
> > discussion AT sctp.de
> > Subject: [SCTP-Discussion] About the SCTP checksum
> >
> > Hi all,
> > I have a question about the SCTP checksum.
> > The RFC defines a single checksum covering the entire packet.
> > This could lead to inefficient behaviour in the case of data
> > corruption, 'cause an error in a single DATA CHUNK results in the
> > retransmission
> > of all other DATA CHUNKS eventually boundled in the same packet.
> > This, combined whit the need to re-compute the Adler-32 checksum 4 the
> > entire
> > packet,
> > is an overhead, I think.
> > Do You know the reason of this choice, or what I missunderstand ??
> > Thank's.
> >
> >
> > _______________________________________________
> > discussion mailing list
> > discussion AT sctp.de
> > http://www.sctp.de/mailman/listinfo/discussion
> _______________________________________________
> discussion mailing list
> discussion AT sctp.de
> http://www.sctp.de/mailman/listinfo/discussion





Archive powered by MHonArc 2.6.19+.

Top of page