General Switch Management Protocol (gsmp)

Sanjay Wadhwa | 14 Feb 2005 23:08

GSMPv3 & L2C

Section 3.1.4 in the GSMPv3 draft describes extension TLVs. There is a restriction of "at most one
extension TLV per message type". Is there a particular reason for this restriction ? There might be cases
where a particular technology needs more than one such (optional) TLV in a message. One option could be to
have the "extension value" be composed of multiple sub-TLVs (with their own well defined format). But, it
is simpler to allow multiple top level TLVs. Based on the "length" in the GSMP message header and the
"length" in each TLV, the parsing is not a problem. For sanity check number of TLVs in the message can be
included in the message. The device that does not understand a particular TLV can silently ignore the TLV
and continue on.
In context of L2C this is needed for a BRAS (controller) to configure line characterstics of a DSL line on a
DSLAM. There is more than one parameter that can be configured (multiple, optional TLVs in the port
management message seems like an appropriate way to implement this functionality). Any thoughts on not
having the restriction of a single "extension" TLV per message type ?

Thanks
-Sanjay

headers

avri doria | 10 Mar 2005 16:36

Re: GSMPv3 & L2C

Hi,

Sorry I did not answer this sooner.  My inclination would be to 
maintain a single extension TLV per technology type and allow the 
technology specific protocol extension to nest sub-TLVs within that 
structure.

My reasons are that I think it would be simpler for the base protocol 
implementation and would organize the IANA effort for defining new TLV 
fields with more flexibility for the extension writer; i.e each 
technology specific extension defines the fields it needs and does its 
own IANA registrations.

My goal with the base has to keep it as simple as possible so that 
hoepfully once it get back on the standards track, it can remain 
unchanged while the technology specific efforts continue independently.

I would be interested in hearing other people on this list comment on 
this issue.

thanks
a.
On 14 feb 2005, at 16.08, Sanjay Wadhwa wrote:

> Section 3.1.4 in the GSMPv3 draft describes extension TLVs. There is a 
> restriction of "at most one extension TLV per message type". Is there 
> a particular reason for this restriction ? There might be cases where 
> a particular technology needs more than one such (optional) TLV in a 
> message. One option could be to have the "extension value" be composed 
> of multiple sub-TLVs (with their own well defined format). But, it is

(Continue reading)