>>>>> On Mon, 03 Dec 2001 14:24:23 +0900, 
>>>>> JINMEI Tatuya <[EMAIL PROTECTED]> said:

> An alternative candidate is to allow the stack to pass ancillary data
> items even without TCP data (recvmsg() would then return 0 in such a
> case.)  With this approach, however, the application may not be able
> to determine the termination of a connection.  The traditional coding
> style like

>       if ((n = recvmsg(s, &msghdr, 0)) == 0)
>               return(-1);     /* end of the connection */

I've considered a bit more about this issue.  Here is a comparison
between those two approaches with pros and cons list:

Pros and cons of using the (new) recvmsg() approach:
  pro: easy migration from 2292bis-02 (and before) applications
  pro: less impact on existing 2292bis-02 (and before) kernels

  con: hard migration from RFC 2292 applications
  con: impact on existing RFC 2292 kernels
  con: impact on the traditional socket semantics
       e.g. need additional condition to detect an end of connection (above)
       e.g. it is not clear what if the app calls shutdown(0)
            (i.e. makes the socket unreadable)?  Should the reception
            of ancillary data also be disabled?

Pros and cons of using the RFC2292-style socket option (described in
the 03 draft):
  pro: easy migration from RFC 2292 applications
  pro: less impact on existing RFC 2292 kernels
  pro: less impact on the traditional socket semantics

  con: hard migration from rfc2292bis-02 (and before) applications
  con: impact on existing rfc2292bis-02 (and before) kernels

So, apart from the impact on the socket semantics, this is a migration
(or backward compatibility) issue.  I'd like to know the current
situation about the deployment of rfc2292bis-02 (or before), and
opinions of implementors that have implemented or have a plan to
implement RFC2292 and/or rfc2292bis.

                                        JINMEI, Tatuya
                                        Communication Platform Lab.
                                        Corporate R&D Center, Toshiba Corp.
                                        [EMAIL PROTECTED]
--------------------------------------------------------------------
IETF IPng Working Group Mailing List
IPng Home Page:                      http://playground.sun.com/ipng
FTP archive:                      ftp://playground.sun.com/pub/ipng
Direct all administrative requests to [EMAIL PROTECTED]
--------------------------------------------------------------------
  • ... Internet-Drafts
    • ... Vladislav Yasevich
      • ... JINMEI Tatuya / 神明達哉
        • ... Lilian Fernandes
          • ... JINMEI Tatuya / 神明達哉
    • ... Tim Hartrick
      • ... JINMEI Tatuya / 神明達哉
        • ... JINMEI Tatuya / 神明達哉
          • ... JINMEI Tatuya / 神明達哉
            • ... Vlad Yasevich
              • ... JINMEI Tatuya / ソタフタテ」コネ
                • ... Vladislav Yasevich
                • ... JINMEI Tatuya / 神明達哉
            • ... Atsushi Onoe

Reply via email to