Re: [Diffserv] APIs for diffserv

[Brian E Carpenter <[e-mail]>]

Thomas Narten wrote:
> 
> The IPv6 basic and advanced APIs include placeholders/hooks for
> specifying the Flow Label on outgoing IPv6 packets. But the way the
> API is specified, the entire first 32 bits of the IP header can be
> specified, which includes the Traffic Class field. Many of the details
> are unspecified, however.
> 
> Questions:
> 
> 1) What work has been done with regards to specifying an API for
>    setting the Traffic Class fields? Anything other than the IPv6
>    APIs?

There was draft-itojun-ipv6-flowlabel-api-01.txt (April 2001).
itojun may want to comment.
> 
> 2) Is it adequiate for the existing IPv6 APIs (basic and advanced)
>    to simply allow the traffic class to be set via the flow label
>    mechanism (without specifying the details much)?

It needs to be possible to set the 6 bits of the DSCP independently
of either the flow label or the ECN bits. Ditto the flow label. 
If the proposed APIs can't do that, they are inadequate.

> 
> 3) What about APIs for IPv4? Aren't they needed too?

I thought that you could set the TOS byte in the v4 API. Again,
you have to avoid messing up the ECN bits, but I would expect the
TCP/IP stack to take over those bits.

> 
> 4) Or is there an assumption here that endpoint applications setting
>    the traffic class themselves via an API is not particularly
>    interesting or necessary?

It's been debated, but in the general case it is desirable. The DSCP bits
may be set within the stack by a QOS manager, or downstream by a router,
but apps should have the ability in principle. The flow label may be set
within the stack, or in principle by the app.

   Brian

> 
> Have a look at
> 
>     draft-ietf-ipngwg-rfc2553bis-07.txt and
>     draft-ietf-ipngwg-rfc2292bis-07.txt
> 
> for the details of what has been specified so far.
> 
> Thoughts of what (if anything more) should be done with APIs for
> diffserv would be appreciated.
> 
> Thomas
> _______________________________________________
> diffserv mailing list
> [e-mail]
> https://www1.ietf.org/mailman/listinfo/diffserv
> Archive: http://www.ietf.org/mail-archive/working-groups/diffserv/current/maillist.html

-- 
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
Brian E Carpenter 
Distinguished Engineer, Internet Standards & Technology, IBM 
On assignment at the IBM Zurich Laboratory, Switzerland
_______________________________________________
diffserv mailing list
[e-mail]
https://www1.ietf.org/mailman/listinfo/diffserv
Archive: http://www.ietf.org/mail-archive/working-groups/diffserv/current/maillist.html