[Forces-protocol] Re: Instance Select

["Weiming Wang" <[e-mail]>]

Hi Joel,

----- Original Message -----
From: "Joel M. Halpern" <[e-mail]>
Subject: Re: Instance Select


> While I am concerned about feature creep in adding this multi-targetting, I
> will stop objecting to it at this point.
Could you show why and what's the benifits to defer it. What I see the issue is
it is more a basic one than a value-added one, for it is quite  like a printer
which can only print one page at a time if without multiple selecting.

> If we want to add such, there is no need to use a TLV to represent
> it.  Simply define the instance ID encoding so we can tell ranges from
> individual instances.
Does it mean the coding as:

LFBCLASSID InstanceID <RangeMark | Instance ID >+  <OPER>+ ?

If is, I'm afraid there is a size problem for it, for the InstanceID number
changes and we can not tell an InstanceID with the OPER TLV type field.
If it means using some tag in InstanceID field to indicate the InstanceID and
the Range mark, then I think it will still limit the OPER TLV type definition.
Could you show more on this?

>Using a TLV for information whose presence
The presense is definite, for we at least need one Instance ID.

>and
> position is mandatory does not add much, and can in fact cause problems.
Joel, I know you are quite experienced and I'm really very much interested in
what you think of the problems here. From my limited experience, I think TLV
followed by TLV approach is a very common PDU scheme for protocol design,  and
will lead to no problems that I can think of. Actually, in our current scheme, I
think we have already used LFBselectTLV followed by many other LFBselectTLV
format. And I also see we will use Path-Data TLV quene inevitably. Maybe I'v
missed something.

Thank you.
Weiming
>
> Yours,
> Joel
>
> At 12:48 PM 10/21/2004 +0800, Wang,Weiming wrote:
> >Hi Jamal, Hormuzd, etc,
> >
> >To summarize the discussions on multiple instances, I try to propose
following
> >scheme for instance selection, which follows Robert's idea and Jamal's
format,
> >as:
> >
> >PL level PDU : = MAINHDR<LFBselect>+
> >LFBselect := LFBCLASSID InsSelect <OPER>+
> >InsSelect := InstanceID <RangeMark | Instance ID >+
> >RangeMark := '0xFFFFFFFF'; the value is the same as Broadcast Instance
> >address,
> >no worry of ambiguity here.
> >
> >The InsSelect is a TLV, whose structure is shown as:
> >
> >main hdr (eg type = config)
> >      |
> >      |
> >      +-- T = LFBselect
> >      |        |
> >      |        +- LFBCLASSID = target LFB class
> >      |        |
> >      |        |
> >      |        +- T = InsSelect
> >      |        |   |
> >      |        |   V = InstanceID <RangeMark | Instance ID >+
> >      |        |
> >      |        +- T = operation { ADD, DEL, GET, etc}
> >      ...
> >
> >Best Regards,
> >Weiming
>
>



_______________________________________________
Forces-protocol mailing list
[e-mail]
https://www1.ietf.org/mailman/listinfo/forces-protocol