Re: question on draft-lfblb-draft.txt?
Jia Fenggen <[email protected]>
| Newsgroups | gmane.ietf.forces |
|---|---|
| Message-ID | <[email protected]> |
I agree that Port LFB should only include functionality that is closely coupled to the media.I take functionality coupled to the media as assembling cells to packets and deassembling packets to cells,media data form coversion to FE processing data form,that is to say the port LFB FE side input and output is frame,the media side output and input is cell or some form else,and I temp to add to LFB as L2 decap and L2 encap to capture the frame processing,like classify frames into ipv4 and ipv6,strip off L2 header and do L2 filtering,what do you suppose? Yours,Fenggen >From: "Joel M. Halpern" <[email protected]> >Reply-To: "Joel M. Halpern" <[email protected]> >To: [email protected] >Subject: Re: question on draft-lfblb-draft.txt? >Date: Fri, 22 Sep 2006 09:24:28 -0400 > >I have removed the diagam below. It ended up so wide that I could >not get a good view of it. > >I am pretty sure that you are asking what the relationship is >between the port LFB and various forms of processing such as layer 2 >validation, layer 3 validation, forwarding lookup, etc. > >The general guideline is that processing that is logically distinct >should be in different LFBs. Thus L3 validation L3 Next Hop >selection, and Next hop application are ll proposed in the library >to be separate LFBs. >The Port LFB should only include functionality that is closely >coupled to the media. So I have suggested that media decapsulation >(or encapsulation) goes in the Port (or Generic Connectivity) LFB. >Media errors are also counted in that LFB. > >However, that decomposition is just a suggestion. The working group >has not adopted the library draft. Other folks haven't started >writing class definitions for the draft. If there is a different >decomposition that helps us represent the range of implementations >of routers and other packet processing devices, I am open to hearing >it. > >Yours, >Joel M. Halpern > >At 01:11 AM 9/22/2006, Jia Fenggen wrote: >>Still another question about port LFB,in convention router,packet >>processing on interface always be divided into to stages:ingress >>and egress processing,ingress handles things like L2 >>decap,Forwarding table lookup and then send the packets to fabric >>which will deliver it to egress side for queuing and scheduling for >>sending out to media.This two stage may reside in an interface >>processing card in modern routers.My question is the Port LFB >>reside and fuctions in such model.As model pictures: >... >>what's relationship between Ingress\Egress in the above draw and >>the Port LFB?Should we give an example LFB topology which use the >>Port LFB,in part to satify the ForCES Requriment and in another >>aspect make the reader known more of the model? Yours,Fenggen _________________________________________________________________ 璐逛杞 MSN Explorer: http://explorer.msn.com/lccn/