[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: LMP & neighbor discovery
Martin,
> We have to be aware that LMP is a generic protocol not tailored
> specifically for SONET/SDH. It is very dangerous to start adding
> protocol specific concepts in the spec.
The neighbor discovery description in
http://ops.ietf.org/lists/ccamp/ccamp.2002/msg00750.html
is valid for any technique that allows for sending and reading
an in-band "access point identifier" that can contain at least
15 T.50 characters. I.e. it is at least applicable to SONET/SDH
and OTN. It is also applicable to transparant optical cross
connects that are capable to generate a "test signal".
This makes the description general enough - in my mind - to
include it as an optional procedure in the LMP draft.
We might indeed want to remove the first byte (the CRC) from
the described format of the access point identifier. This
leaves us with 15 T.50 characters and allows the addition of
the CRC to be specific for the protocol in the transmission
layer (the layer of the data links).
> To represent TTPs and CTPs, one can use the generic concepts
> of ports and component links (see Introduction for a definition
> of port and component link). There are already a number of
> Internet drafts that use these concepts.
Please let me know how you would describe - with these terms -
the point I made in
http://ops.ietf.org/lists/ccamp/ccamp.2002/msg00651.html
Best regards,
Michiel
Martin Dubuc wrote:
>
> Zhi-Wei,
>
> We have to be aware that LMP is a generic protocol not tailored specifically for SONET/SDH. It is very dangerous to start adding protocol specific concepts in the spec.
>
> To represent TTPs and CTPs, one can use the generic concepts of ports and component links (see Introduction for a definition of port and component link). There are already a number of Internet drafts that use these concepts.
>
> Switching to CTPs and TTPs at this point would be very dangerous in that it would require changes in several drafts and would also alienate implementations that are not SONET centric.
>
> LMP is very closely linked with GMPLS. I am not sure why you single-handedly discard the GMPLS argument for adequacy.
>
> I still believe that the current draft allows vendors to implement automatic discovery. The link connectivity verification described in the draft can be used to fulfill this requirement. There is no need for extension of the messages to support this.
>
> Martin
>
> <...snip...>