Mahesh,
Your review brings up interesting aspects.
I can provide some initial responses,
Den 2017-03-07 kl. 03:47, skrev Mahesh Jethanandani:
Reviewer: Mahesh Jethanandani
Review result: Has Nits
I have reviewed this document as part of the Operational
directorate’s ongoing effort to review all IETF documents being
processed by the IESG. These comments were written with the intent of
improving the operational aspects of the IETF drafts. Comments that
are not addressed in last call may be
included in AD reviews during the IESG review. Document editors
and WG chairs should treat these comments just like any other last
call comments.
Document reviewed: draft-ietf-slim-negotiating-human-language-08
Status:
Ready with comments.
Summary:
This document adds new SDP media-level attributes so that when
establishing interactive communication sessions ("calls"), it is
possible to negotiate (communicate and match) the caller's language
and media needs with the capabilities of the called party.
The document is short and easy to read. And it seems to have
considered many aspects of trying to negotiate a common human language
or capability. This review looks at the document more from a operator
or management perspective.
Operational considerations:
>From a operations perspective, there may be a need to troubleshoot the
interface that sets up the negotiated human language. Identifying
consistent methods of information that should be counted by both
parties will go a long way in debugging a problem. For example, in
this case, it would be helpful to start by collecting how many
requests were made, how many found a language or medium in common and
how many were rejected because a common match was not found.
Management considerations:
The old adage says, “Anything that can be configured, can also be
misconfigured”, unless that is somehow made less possible by providing
default values, modes or parameters. This can be something that can be
defined using a data model in YANG.
I assume that the default behavior of receiving a SDP attribute that
one does not support, results in a throw away of that particular
attribute, and not the whole message, if combined with other
attributes. Is this documented somewhere? If not, what does the
deployment scenario look like, particularly with existing solutions?
In section 5 of RFC 4566 the SDP specification, it is said:
"An SDP parser MUST ignore any attribute it doesn't understand."
In a real world, the deployment will be gradual, so many times, either
the caller UA or the callee UA will not understand these attributes.
Some possible default actions are mentioned in the draft for cases when
there are no settings or no support in either end, but the draft is not
intended to provide exact solutions for the negotiation, but rather only
provide a protocol for conveying indications sufficient for performing
the negotiation.
What is the impact on network operations if for example either the
translator or relay agent fails? How would that impact the
negotiation?
The draft does not intend to dictate how and by what party a
language/modality gap is detected, so that a translating agent or
modality conversion relay service will be desired and invoked in the
call. The intention is to make such detection and invocation possible,
at least regarding relay services. The natural way of using the
mechanism of the draft is that each party involved in the call setup
will modify the hlang attributes to indicate the joint capabilities and
preferences of the so far involved parties. If invocation of a party
fails, its modification of the hlang attributes does not take place and
the answer will then not likely contain a satisfying set of languages
for the offering party. This may be a case when the paranthesis in this
sentence in section 5.2 comes into play: "In an answer, 'hlang-send' is
the language the answerer will send
when using the media (which in most cases is one of the languages in
the offer's 'hlang-recv'), and 'hlang-recv' is the language the
answerer expects to receive in the media (which in most cases is one
of the languages in the offer's 'hlang-send')."
If the answering party was the one who intended to invoke the
translation or relaying if needed, then it can also deny the call if
that is indicated as a preference in the incoming call.
These details belong to the negotiation mechanism that was not intended
to be described in detail in the draft.
Also, what is the test, both active and passive for the correct
operation? Is there a counter being maintained for both correct and
incorrect negotiation. Goes back to the question of what counters are
being maintained. Such counters should include values that enable
isolation of faults. For example, if negotiation fails, what are the
more specific counters that isolate what within the negotiation
failed?
Fault Management:
In addition to collection information on how the negotiation is
working, it is important to be able to propagate both fault and health
indicators to a management application. Such information needs to be
documented.
Accounting Management:
Finally, it is always helpful to collect information on utilization
from capacity, trend analysis, cost allocation, auditing and billing
perspective.
Interesting aspects. Let us see what others say about what we can
include at this stage and what should be brougth up in follow-up documents.
/Gunnar
Idnits:
A run of idnits came out clean.
_______________________________________________
SLIM mailing list
SLIM@xxxxxxxx
https://www.ietf.org/mailman/listinfo/slim
--
-----------------------------------------
Gunnar Hellström
Omnitor
gunnar.hellstrom@xxxxxxxxxx
+46 708 204 288