Re: [trill] Adam Roach's No Objection on draft-ietf-trill-address-flush-05: (with COMMENT)

[Donald Eastlake <[e-mail]>]

Hi Adam,

On Wed, Feb 7, 2018 at 11:30 PM, Adam Roach <[e-mail]> wrote:
> Adam Roach has entered the following ballot position for
> draft-ietf-trill-address-flush-05: No Objection
>
> ...
>
> ----------------------------------------------------------------------
> COMMENT:
> ----------------------------------------------------------------------
>
> Thanks to everyone who contributed to writing this document.
>
> I'm concerned that the interaction between the various extensible Address Flush
> message TLVs isn't very clearly spelled out. As far as I can tell, the text that
> attempts to describe the interactions is:
>
>>     If the set of MAC addresses accumulated from parsing the address
>>  flush message is null, the message applies to all MAC addresses.
>>     If the flag indicating the presence of an All Data Labels TLV is
>>  true, then the address flush message applies to all Data Labels and
>>  the set of Data Labels and block of Data labels specified has no
>>  effect. If the flag indicating the presence of an All Data Labels TLV
>>  is false, then the address flush messages applies only to the set of
>>  Data Labels accumulated from parsing the message; if that set is
>>  null, the address flush message does nothing.

The TLVs are all parsed accumulated a possibly null set of Data
Labels, an "All Data Labels flag", and a possibly null set of MAC
addresses.

> Based on this (and the fact that their implementation is optional), I infer
> that the MAC address TLVs are intended to further restrict the addresses
> indicated by TLV types 1 through 5, rather than expand upon them.

That's right.

> I'm less
> sure about whether they have any impact on Type 6. I would expect that they
> do, but the text above ("applies to all Data Labels") kind of sounds like they
> don't.

Type 6 only affects the Data Labels that are being flushed. It has no
effect on the set of MAC addresses which set defaults to "all".

> What would seem to make sense here (inasmuch as it provides maximal flexibility)
> is:
>
> if (TLV7 ∪ TLV8 = {})
>   Addresses to Flush = (TLV1 ∪ TLV2 ∪ TLV3 ∪ TLV4 ∪ TLV5 ∪ TLV6)
> else
>   Addresses to Flush = (TLV1 ∪ TLV2 ∪ TLV3 ∪ TLV4 ∪ TLV5 ∪ TLV6) ∩ (TLV7 ∪ TLV8)

That's correct but not complete. If the set of Data Labels indicated
by any of TLV1 though TLV5 that are present is null, the the Address
Flush applies to the Data Label after the TRILL Header and Inner MAC
addresses. It also seems quite confusing to have TLV7 and 8 as if they
were the same sort of thing as the other TLVs when they are MAC
addresses and the others all relate to .

Every entry that is flushed consists of a nickname, a Data Label, and
a MAC address. Nicknames are handled the same way for both the VLAN
Block only case and the extensible case. The complete set of entries
that is flushed is the cross product of a set of nicknames, a set of
Data Labels, and a set of MAC addresses, where the set of Data Labels
and/or MAC addresses can be "all".

> If that's the intention, I think the normative explanation needs to be clearer.
> If that's not the intention, I sill think the normative explanation needs to be
> clearer.

We'll see what we can do to clarify.

> My remaining comments are editorial.
> ---------------------------------------------------------------------------

I'm fine with the changes below and have provided and explanation of "RB1".

Thanks,
Donald
===============================
 Donald E. Eastlake 3rd   [phone] (cell)
 155 Beaver Street, Milford, MA 01757 USA
 [e-mail]

> Please expand the following acronyms upon first use and in the title;
> see https://www.rfc-editor.org/materials/abbrev.expansion.txt for guidance.
>
>  - TRILL
>  - TC
>  - TCN
>  - MSTP
>
> While the following terms are defined in cited documents, you may wish to also
> consider expanding them in this document's acronym list for the convenience of
> the reader:
>
>  - MAC
>  - FCS
>
> Finally, please explain the use of "RB1" in the Introduction.

It's use in this case is pretty much non-existent... :-) A typical
style in TRILL document is to say "A TRILL switch RB1 ..." Then later
you can just reference RB1 rather than having to talk about it in some
more complex say. (Usually there is also an RB2, RB3, ... but not
always.) Since the "RB1" is never referenced again in this case, it
can just be deleted.

> ---------------------------------------------------------------------------
>
> §1:
>
>>  Another example is based on Appendix A.3 of [RFC6325] ("Wiring Closet
>>  Topology") presents a bridged LAN connected to a TRILL network via
>>  multiple RBridge ports.
>
> This should be either:
>
>    Another example, based on Appendix A.3 of [RFC6325] ("Wiring Closet
>    Topology"), presents a bridged LAN connected to a TRILL network via
>    multiple RBridge ports.
>
> or...
>
>    Another example is based on Appendix A.3 of [RFC6325] ("Wiring Closet
>    Topology"), which presents a bridged LAN connected to a TRILL network
>    via multiple RBridge ports.
>
> ---------------------------------------------------------------------------
>
> §1.1:
>
>>  The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT",
>>  "SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this
>>  document are to be interpreted as described in [RFC2119].
>
> This document makes use of lowercase versions of these terms as well; please
> consider using the RFC 8174 boilerplate.
>
> ---------------------------------------------------------------------------
>
> §2.2:
>
>>  VLANs/FGLs if it occurs in any TLV in the address flush message. A
>>  MAC addresses might be indicated more than once due to overlapping
>
> "A MAC address..." or "MAC addresses..."
>
>
> ---------------------------------------------------------------------------
>
> §2.2:
>
>>  MAC addresses if it occurs in any TLV in the address flush message.
>>     If the set of MAC addresses accumulated from parsing the address
>>  flush message is null, the message applies to all MAC addresses.
>>     If the flag indicating the presence of an All Data Labels TLV is
>>  true, then the address flush message applies to all Data Labels and
>
> The staggered indenting here looks a bit odd. Were these intended to be a
> bulleted list?