Hi Itojun,
I am not a kernel specialist, but I assume that from
a performance point of view it would be much more performant
to have several flows using only one socket instead of
having a single socket for each flow?
An other concern is the redundancy of information in
ports and labels.
Cheers,
Jochen
------------------------------------------------------------------
Jochen Metzler Siemens AG
Lise-Meitner-Stra�e 5, 89081 Ulm
Phone: +49.(0)731.9533.390
Mobile Networks Fax : +49.(0)731.9533.336
ICM N MR UR SE 2 mailto: [EMAIL PROTECTED]
-----Urspr�ngliche Nachricht-----
Von: [EMAIL PROTECTED] [SMTP:[EMAIL PROTECTED]]
Gesendet am: Donnerstag, 30. November 2000 11:33
An: Metzler Jochen
Cc: 'Elwyn Davies'; '[EMAIL PROTECTED]'
Betreff: Re: AW: Usage of IPv6 flow label
>Hi,
>I fully agree with Itojun's draft on generating the flow label within the
network code. But
>I see no reason, why it should not be (at least) readable from application
layer and why
>a single UDP or TCP session should not handle more than one streams.
I admit there could be people who wants to use:
- multiple flows (= multiple flow label value) in a single TCP session
- multiple TCP sessions with single flow label value
however, I could not find a good (a simple and usable) UNIX API to
make
the above option happen, so I simplified the story.
At least, the latter case can easily be handled by "single TCP session
per single flow label".
I documented the issue in section 5 of my draft.
itojun
--------------------------------------------------------------------
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]
--------------------------------------------------------------------
