Loading ...
Sorry, an error occurred while loading the content.

[decentralization] SCTP?

Expand Messages
  • Zane Thomas
    All, In the past there have been discussions of the relative merits of TPC & UDP for P2P networks. However I don t recall any discussion about the possible
    Message 1 of 7 , Jul 29, 2001
    • 0 Attachment
      All,

      In the past there have been discussions of the relative merits of TPC & UDP
      for P2P networks. However I don't recall any discussion about the possible
      use of SCTP (http://www.ietf.org/rfc/rfc2960.txt) - do any of you have
      thoughts to share on the possible use of this protocol?

      Zane
    • Zane Thomas
      Coderman, ... still high (like ... Seems to me that unless you can live with UDP as-is there is always going to be peer-connection overhead - the question is:
      Message 2 of 7 , Jul 29, 2001
      • 0 Attachment
        Coderman,

        > Looks pretty sweet. My only complaint is that per connection overhead is
        still high (like
        > TCP).

        Seems to me that unless you can live with UDP as-is there is always going to
        be peer-connection overhead - the question is: what do you get for it?

        > Do you know of any implemented stacks for this protocol?

        See http://www.sctp.org/

        Zane
      • coderman
        Looks pretty sweet. My only complaint is that per connection overhead is still high (like TCP). Do you know of any implemented stacks for this protocol?
        Message 3 of 7 , Jul 29, 2001
        • 0 Attachment
          Looks pretty sweet. My only complaint is that per connection overhead is still high (like
          TCP).

          Do you know of any implemented stacks for this protocol?



          Zane Thomas wrote:
          >
          > All,
          >
          > In the past there have been discussions of the relative merits of TPC & UDP
          > for P2P networks. However I don't recall any discussion about the possible
          > use of SCTP (http://www.ietf.org/rfc/rfc2960.txt) - do any of you have
          > thoughts to share on the possible use of this protocol?
          >
          > Zane
          >
          > To unsubscribe from this group, send an email to:
          > decentralization-unsubscribe@egroups.com
          >
          >
          >
          > Your use of Yahoo! Groups is subject to http://docs.yahoo.com/info/terms/
        • Bram Cohen
          ... That looks like good research, but not good enough to warrant not using TCP. The rule for what low-level protocol you need to use is to ask yourself Does
          Message 4 of 7 , Jul 29, 2001
          • 0 Attachment
            On Sun, 29 Jul 2001, Zane Thomas wrote:

            > See http://www.sctp.org/
            >

            That looks like good research, but not good enough to warrant not using
            TCP.

            The rule for what low-level protocol you need to use is to ask yourself
            'Does it need to be lossy?' If the answer is yes use UDP, otherwise use
            TCP.

            -Bram Cohen

            "Markets can remain irrational longer than you can remain solvent"
            -- John Maynard Keynes
          • Zane Thomas
            Bram, ... Rules were made to be broken, good rules were made such that people don t want to break them very often. :-) Seriously, I recall someone here has
            Message 5 of 7 , Jul 30, 2001
            • 0 Attachment
              Bram,

              > The rule ...

              Rules were made to be broken, good rules were made such that people don't
              want to break them very often. :-) Seriously, I recall someone here has
              built a reliable transport over top of UDP. There may be cases where that's
              a good idea.

              Zane
            • Abhik Shah
              Hi, I ve only given the rfc a cursory scan but it looks like sctp would involve at least the same amount of overhead as TCP. Usually, the argument against TCP
              Message 6 of 7 , Jul 30, 2001
              • 0 Attachment
                Hi,
                I've only given the rfc a cursory scan but it looks
                like sctp would involve at least the same amount of
                overhead as TCP. Usually, the argument against TCP
                for P2P applications is the extra overhead and I don't
                see SCTP solving that.

                On a related note, has anyone checked out T/TCP
                (rfc1379) for P2P apps? T/TCP is an extension to TCP
                that allows for lower overhead (simpler connection
                build and shutdown -- no 3-way handshake) on certain
                connections. Basically, if you have communicated with
                a host recently, you can skip the expensive connection
                build and shutdown. I think the savings is about 1
                RTT for a 1 segment request/response connection. It
                still retains all of TCP's features (relability, flow
                control, ordering, etc)

                $.02,
                abhik.

                --- Zane Thomas <zane@...> wrote:
                > All,
                >
                > In the past there have been discussions of the
                > relative merits of TPC & UDP
                > for P2P networks. However I don't recall any
                > discussion about the possible
                > use of SCTP (http://www.ietf.org/rfc/rfc2960.txt) -
                > do any of you have
                > thoughts to share on the possible use of this
                > protocol?
                >
                > Zane
                >
                >
                > To unsubscribe from this group, send an email to:
                > decentralization-unsubscribe@egroups.com
                >
                >
                >
                > Your use of Yahoo! Groups is subject to
                > http://docs.yahoo.com/info/terms/
                >
                >


                __________________________________________________
                Do You Yahoo!?
                Make international calls for as low as $.04/minute with Yahoo! Messenger
                http://phonecard.yahoo.com/
              • coderman
                I implemented a reliable UDP transport for a project I am working on. However, there is a bit more to this protocol. It basically provides for unreliable, or
                Message 7 of 7 , Jul 30, 2001
                • 0 Attachment
                  I implemented a reliable UDP transport for a project I am working on. However, there is a
                  bit more to this protocol. It basically provides for unreliable, or reliable transport of
                  datagrams (no sliding windows, no flow control (well, no incoming flow control), that is
                  required at higher protocol levels).

                  It also provides multiplexing of large numbers of logical connections (hundreds of
                  thousands) with very little per connection overhead.

                  Anyway, like you said, this was for a very specific implementation need which UDP and TCP
                  did not fill. SCTP comes closer, but its overhead is still far to high for what I am
                  using this other UDP based transport for.

                  Best regards...


                  Zane Thomas wrote:
                  >
                  > Rules were made to be broken, good rules were made such that people don't
                  > want to break them very often. :-) Seriously, I recall someone here has
                  > built a reliable transport over top of UDP. There may be cases where that's
                  > a good idea.
                  >
                Your message has been successfully submitted and would be delivered to recipients shortly.