Re: ForCES Protocol Implementation Issues

"Wang,Weiming" <wmwang2001@hotmail.com> Wed, 19 September 2007 06:45 UTC

Message-Id: <WED.19.SEP.2007.144533.0800.>
Date: Wed, 19 Sep 2007 14:45:33 +0800
From: "Wang,Weiming" <wmwang2001@hotmail.com>
Subject: Re: ForCES Protocol Implementation Issues
Comments: cc: Wang Weiming <wmwang@mail.zjgsu.edu.cn>
MIME-Version: 1.0
Content-Type: text/plain; charset="gb2312"
Content-Transfer-Encoding: base64

Jamal, and all,

I'm always awake for ForCES :) Only a few days late, sorry.

Time elapsed so quickly. I just remembered years ago we had discussed the issue. I cann't very clearly remember the results, only have some thoughts remained in mind.

I remember at first, I was not very in favor of the fig 4 FE State machine, where FE admin UP/DOWN was first introduced but without further clear explaination. When this issue was raised, the final consensus seemed to be that the FE admin UP/DOWN attribute would be added in the FEO. Along with the FEO moved to the model draft, I did not follow the issue very closely from then. 

Now the question seemed to be that the model draft only defined the FE status as a read-only attribute, which makes the CE set impractical. 

I guess the reason the author of model draft did not set the FE status as writable attribute is that this FE UP/DOWN operation is still unclear for implementation, i.e., what should in reality be done when FEO get this setting?

I recalled we ever discussed that every LFB should define a On/Off status so as to associate it with the FE UP/DOWN operation, i.e., traslate FE UP/Down setting as all existing LFB On/Off setting. I'm not sure if this thought can help in some way.

thanks,
Weiming


----- Original Message ----- 
From: "Jamal Hadi Salim" <hadi@znyx.com>

> On Mon, 2007-17-09 at 06:57 -0400, Joel M. Halpern wrote:
> > As far as I can tell, this (turning an FE on and off) is not a 
> > problem that needs solving.
> 
> The last time we had this discussion there there were strong objections
> - in particular from Weiming (CCed to wake him up). Our implementation
> doesnt use this feature (and does what you described). The idea is to be
> able to populate LFB tables before allowing the FE to process packets.
> 
> Your earlier suggestion to have a read-write admin-status and a
> read-only operation-status would be the most sensible. 
> The discussion as i recall was to not have adminEnable enumeration
> because operational enable in this case implied Admin enabled.
> 
> cheers,
> jamal
>