[Insipid] Review of draft-ietf-insipid-logme-marking-08

[Christer Holmberg <[e-mail]>]

Hi,

I have reviewed draft-ietf-insipid-logme-marking-08.txt.

Summary:
--------

In general, the draft looks good, and while my comments are mostly
editorial I think there are technical procedures that need to be clarified.


Section 3.2.:
-------------

The text is very unclear, especially regarding the scope (single request
vs dialog) of a marking.

In the beginning of the section, you should say something like:

- If the marker is included in a dialog-initiation request, the scope of
the logging applies to the lifetime of the dialog, and each request and
response. 
- If the marker is included in a mid-dialog request, or in a response to a
mid-dialog request, the scope of the logging only applies to the request
(or the response). 

In case of dialog wide logging, where the marker is included in the
initial INVITE, does one still have to include the marker in all
sub-sequent mid-dialog messages? The examples in section 4, and the text
in section 5.1 seem to imply so. If so, I think it should be explicitly
indicated.

The title talks about ³stopping² logging, but there are no procedures for
doing that. When reading section 5.1., it seems like not including the
marker in a mid-dialog message, even if the marking has been negotiated
for the whole dialog, would stop the marking. However, when reading
section 5.1.1 it seems like not including the marker in a mid-dialog
message is an error. It needs to be explicitly clarified whether a
dialog-wide logging can be stopped or not, and how.

I suggest to place the 2nd paragraph at the end of the section. It is just
informative text.

Even though it may be obvious, I think it would be good to explicitly say
(e.g., in a Note) that, if the marker is included in a request or
response, it must be included in all re-transmissions of the
request/response. Similarly, if the marker is not included in a
request/response, it must not be included in re-transmissions of that
request/response.


Section 3.6.:
-------------

Is it a normative requirement to log the whole SIP message (I see no MUST)?

Assuming it is a requirement to log the whole SIP message, do we really
want to mandate that? Couldn¹t it be a SHOULD, unless the operator has
configured that only specific parts are to be logged?


Section 3.7.:
-------------

The text says:

 ""Log me" marking is done per-dialog and typically begins at dialog
creation and ends when the dialog ends.²


This only applies to dialog scoped loggings. As indicated in section 3.2,
loggings can also be done per message.


Section 5.1.2.:
---------------

Section 5 is about error cases. Shouldn¹t this be in a separate section?


Section 5.2.:
-------------

See comment as 5.1.2.

In addition, I think this text belongs in section 3.2. I partly suggested
such text already, but I guess it could be more explicit.


Section 7:
----------

In the table, in the Predefined Values column, I suggest to say "No (no
values are allowed)².



Thanks!

Regards,

Christer