The IP over ATM Mailing List Archive by date

Cell Relay Retreat>List Archive>month:1995-Oct> msg00094



[Date Prev][Date Next][Thread Prev][Thread Next]  
  [Date Index][Thread Index][Author Index][Subject Index]

AAL5 CRC and TCP.

  • From: Tim Dwight <0006078043@mcimail.com>
  • Date: Mon, 23 Oct 95 14:29 EST
  • CC: Pat Medved <medved@npr.legent.com>

Pat:

> The AAL5 specification calls for AAL5 to do a CRC for each cell.

No, it doesn't.  AAL5 computes CRC on a CPCS_PDU;  i.e., on an 
encapsulated IP datagram.  Due to IP segmentation an IP datagram
may not equal a TCP segment.  But, it would seem reasonable to assume
that if all IP datagrams are transferred error-free (AAL5 uses a 32-bit
CRC, so it's pretty good at catching bit errors), the TCP segment 
resulting from gluing them back together is also error-free.

> With this being done, and the fact that lines today
> are better than when TCP was first implemented, is there
> still a need to do a checksum in TCP when running over ATM?

Trouble is, how does TCP know it's running over ATM?  IP may know (will
if using RFC-1577, won't if running over LAN Emulation), but we'd need
some hook to pass this to TCP.  Since a device may have both ATM and
non-ATM local interfaces, this information could vary depending on
IP route selection;  even after the TCP connection is established,
the IP layer could re-route from ATM to non-ATM.  I think the semantics
of trying to keep TCP abreast of the current L2, would be difficult.

Best Regards,

Tim