Re: [Slim] I-D Action: draft-ietf-slim-negotiating-human-language-07.txt

[Gunnar Hellström <[e-mail]>]

Den 2017-02-27 kl. 02:01, skrev Randall Gellens:
> At 8:38 AM +0100 2/26/17, Gunnar Hellström wrote:
>
>>  Randall,
>>
>>  One more clarification for issue 12 - the asterisk placement as a 
>> preference hint:
>>
>>  You say:
>>
>>  " it seems better for the answerer to include all media and 
>> languages from the offer that it can support. "
>>
>>  You are right. I misread this paragraph in section 5.2:
>>  " In an answer, 'humintlang-send' is the language the answerer will
>>  send (which in most cases is one of the languages in the offer's
>>  'humintlang-recv'), and 'humintlang-recv' is the language the
>>  answerer expects to receive (which in most cases is one of the
>>  languages in the offer's 'humintlang-send')."
>>
>>  That gave me the impression that I need to answer with only one 
>> language per direction in the whole SDP.
>
> It's per media.
>
>
>>  But in the first paragraph of 5.2, it is said: "to negotiate
>>  which human language is used in each interactive media stream. "
>>
>>  So, we are allowed to include more than one language per direction 
>> in the SDP, but the users are not expected to use more than one 
>> language per direction, at least initially in the call.
>
> There's nothing in the protocol that says anything about people using 
> more than one media in each direction.
>
>
>>  (that makes the wording "will send" and "expects to receive" in the 
>> cited paragraph a bit misleading, because as you point out, some of 
>> them might never be used. "is prepared to send" and "can accept to 
>> receive" might better correspond to what it means. Ending the 
>> sentence with "per media steram" might also help to remind the reader 
>> about the semantics. I leave to you to change these if you want.)
>
> OK, I'm happy to add the additional clarification.  The paragraph now 
> reads:
>
>    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').

Yes, that might be acceptable if "will send when using the media" can be 
read as being conditional: "will send if use of the media for language 
purposes is selected",
but not if it is read as a definite usage "will send for the periods of 
time that this media will definitively be used for language 
communication."  With the latter interpretation we have not improved our 
logic over the lang attribute that could be interpreted as it required 
use of all declared languages in the session.

I still think it is safer to say "is prepared to send" instead of "will 
send".
While for the other direction, I think the "expects to receive" is vague 
enough to mean that another media might be used.

Sorry for continuing to watch out for logic gaps here.

Gunnar
>
>
>
>>  Den 2017-02-26 kl. 07:46, skrev Gunnar Hellström:
>>
>>>  Den 2017-02-25 kl. 20:58, skrev Randall Gellens:
>>>
>>>>  Hi Gunnar,
>>>>
>>>>  At 11:04 AM +0100 2/25/17, Gunnar Hellström wrote:
>>>>
>>>>>  Fine, I find that we have only issues 5, 6 and 12 still to discuss.
>>>>>
>>>>>  You did not answer issue 6, use of asymmetrical language rather 
>>>>> than unidirectional media. I assume you accepted it.
>>>>>
>>>>
>>>>  Yes, I thought I had indicated that, sorry if I didn't.
>>>>
>>>  Good.
>>>
>>>>
>>>>>
>>>>>  On 5, the request to reinsert wording about seeing the speaker in 
>>>>> video, it is still a huge difference in specifying a preference to 
>>>>> see the speaker for language perception reasons, versus only 
>>>>> specifying that I want a video stream for supplementary purposes. 
>>>>> With the current wording in version -07, section 5.3 says that 
>>>>> that combination is undefined. Nothing in the LC discussion 
>>>>> indicated that it should be undefined. Why did you suddenly want 
>>>>> to delete it? It is useful. Please reinsert with the wording 
>>>>> changes I propose.
>>>>>
>>>>
>>>>  The email discussion led me to believe that the text was 
>>>> controversial. We need to get the draft finished, so it's better to 
>>>> delete controversial text than to spend months fighting about it.
>>>>
>>>  The comments were first about the uncertainty about how the "silly 
>>> states" were to be interpreted.
>>>  We described them all but decided to only keep the view of the 
>>> speaker because it is a real and useful case.
>>>  The idea to differentiate spoken and written cases by script tags 
>>> caused discussions and was dropped. The remaining real case with the 
>>> view of the speaker was mentioned twice in the draft, so it was 
>>> recommended that one of them should be deleted, but not both.
>>>
>>>>
>>>>>
>>>>>  On 12, the meaning of the placement of the asterisk, you ask:
>>>>>  "Making the asterisk a purely-advisory hint as to the 
>>>>> least-preferred media/language combination seems harmless enough, 
>>>>> as it would not be required to support it; however, I'm not sure 
>>>>> it provides any benefit: if an offer contains some set of media 
>>>>> with language, and the answerer can support all of them, should 
>>>>> the answerer only include in its answer those without an asterisk? 
>>>>> It seems simpler for the answerer to include everything in the 
>>>>> offer that it can support."
>>>>>
>>>>>  The answering party should aim at answering with one of the 
>>>>> languages that is without the asterisk in the offer. Only if the 
>>>>> answering party does not have capability in a language without an 
>>>>> asterisk, one with asterisk should be selected. Thereby you get 
>>>>> the best opportunity to start the call in a language combination 
>>>>> that satisfies both users.
>>>>>
>>>>>  Example: A hard-of-hearing user can just barely conduct spoken 
>>>>> calls with persons she knows. From others it is much more reliable 
>>>>> to get text. She calls and declares:
>>>>>
>>>>>  m=audio
>>>>>  a=huml-send:en
>>>>>  a=huml-recv:en*
>>>>>  m=text
>>>>>  a=huml-recv:en
>>>>>
>>>>>  The answering party with text capabilities sees that matching 
>>>>> text for sending is higher preferred than talking, and thus responds:
>>>>>
>>>>>  m=audio
>>>>>  a=huml-recv:en
>>>>>  m=text
>>>>>  a=huml-send:en
>>>>>
>>>>>  The answering party sends the initial greeting in text and the 
>>>>> call continues smoothly in well managed langauage/modality 
>>>>> combinations.
>>>>>
>>>>>  Another called party may not have text capabilities, and may 
>>>>> therefore select the less favoured alternative with using speech 
>>>>> both ways, answering:
>>>>>
>>>>>  m=audio
>>>>>  a=huml-recv:en
>>>>>  a=huml-send:en
>>>>>  m=text 0
>>>>>
>>>>>  The answering party starts taking and the parties try as well as 
>>>>> possible to manage the call in this less preferred combination 
>>>>> that may be less reliable.
>>>>>
>>>>>  If the placement of the asterisk had no special meaning as it is 
>>>>> in version -07, it is a high risk that the answering party in the 
>>>>> first example would select to answer with spoken language that 
>>>>> would be unreliably received. Time and effort would be spent by 
>>>>> speech to make the answering party switch to sending text instead 
>>>>> of talking in order to arrange for a more reliable call situation.
>>>>>
>>>>>  If instead the caller only indicated the most favoured combinations,
>>>>>
>>>>>  m=audio
>>>>>  a=huml-send:en
>>>>>  m=text
>>>>>  a=huml-recv:en
>>>>>
>>>>>  Then the answering parties without text capability would not dare 
>>>>> to try to answer, and a reasonably successful call would be missed.
>>>>>
>>>>>  Many other similar realistic examples can be created, where 
>>>>> placement of the asterisk(s) would be a sufficient indication of 
>>>>> lower preference for language match among alternatives that would 
>>>>> make call establishment successful and smooth in many more cases 
>>>>> than without this indication opportunity.
>>>>>
>>>>>  Do you want more examples?
>>>>>
>>>>>  Please accept proposal 12.
>>>>>
>>>>
>>>>  This convinces me that we cannot accept the proposed text, as it 
>>>> would introduce complexity that the WG explicitly decided to not 
>>>> pursue in this draft. In the examples you provided, it seems better 
>>>> for the answerer to include all media and languages from the offer 
>>>> that it can support. This is much simpler, has only trivial 
>>>> drawbacks (extra media negotiated that might not be used), and is 
>>>> what the WG agreed to.
>>>>
>>>  Yes, you could let the answer SDP contain one common language per 
>>> media and direction, but the answering human need guidance on which 
>>> language is best suited to start the conversation. Therefore the 
>>> placement of the asterisk is used to hint the answering party how to 
>>> start the call.
>>>
>>>  The first example above can be modified to:
>>>
>>>  Example: A hard-of-hearing user can just barely conduct spoken 
>>> calls with persons she knows. From others it is much more reliable 
>>> to get text. She calls and declares:
>>>
>>>  m=audio
>>>  a=huml-send:en
>>>  a=huml-recv:en*
>>>  m=text
>>>  a=huml-recv:en
>>>
>>>  The answering party with capabilities for both written and spoken 
>>> English sees that matching text for sending is higher preferred than 
>>> talking and sends the answer indicating the capabilities:
>>>
>>>  m=audio
>>>  a=huml-recv:en
>>>  a=huml-send:en
>>>  m=text
>>>  a=huml-send:en
>>>
>>>  The answering party makes use of the hint that the caller prefers 
>>> to receive written text and therefore sends the initial greeting in 
>>> text and the call continues smoothly in well managed 
>>> langauage/modality combinations.
>>>
>>>  ----------
>>>  There is no complexity left in this solution, it helps to motivate 
>>> why we have the asterisk on media level, and it helps to successful 
>>> call initiations, so I think it should be acceptable.
>>>
>>>  Gunnar
>>>
>>>>
>>>>  --Randy
>>>>
>>>>>  Den 2017-02-25 kl. 01:32, skrev Randall Gellens:
>>>>>
>>>>>>  At 5:35 PM +0100 2/24/17, Gunnar Hellström wrote:
>>>>>>
>>>>>>>  Den 2017-02-23 kl. 05:15, skrev Randall Gellens:
>>>>>>>
>>>>>>>>  Version -07 addresses all comments except for the unresolved 
>>>>>>>> issue of renaming the two attributes which is currently being 
>>>>>>>> discussed on the list, and adding a new attribute for 
>>>>>>>> bidirectionality.
>>>>>>>>
>>>>>>>>  Per Dale's suggestion, the draft adds advice that if a call is 
>>>>>>>> rejected due to no languages in common, SIP response code 488 
>>>>>>>> (Not Acceptable Here) or 606 (Not Acceptable) be used, along 
>>>>>>>> with a Warning header field indicating the supported languages. 
>>>>>>>> The draft registers a new entry in the warn-code sub-registry 
>>>>>>>> of SIP parameters for this purpose. The draft also has an 
>>>>>>>> expanded set of examples.
>>>>>>>>
>>>>>>>  Good progress. Good to see the enriched examples chapter 5.5.
>>>>>>>  I have a few comments on version -07:
>>>>>>>
>>>>>>>  1. Section 4. second line
>>>>>>>  ------------old text----------------------
>>>>>>>  but is not sufficiently sufficiently
>>>>>>>  ------------new text--------------------------
>>>>>>>  but is not sufficiently
>>>>>>>  ----------end of change 1-----------------
>>>>>>>  Motivation: New typo in version -07
>>>>>>>
>>>>>>
>>>>>>  Thanks.
>>>>>>
>>>>>>>
>>>>>>>  2. Section 5.2, first line
>>>>>>>  ----------------old text-----------------
>>>>>>>  This document defines two new media-level ..
>>>>>>>  ----------------new text----------------------
>>>>>>>  This document defines two media-level ...
>>>>>>>  ----------------end of change 2----------------
>>>>>>>  Motivation: It was commented that when the draft is published, 
>>>>>>> this is not new anymore.
>>>>>>>  There are three more occasions of "new" in the document that 
>>>>>>> may be modified as well.
>>>>>>>
>>>>>>
>>>>>>  OK.
>>>>>>
>>>>>>>
>>>>>>>  3. 5.2 second paragraph
>>>>>>>  -------------------old text--------------------------------
>>>>>>>  In an offer, the 'humintlang-send' values indicates the 
>>>>>>> language(s)
>>>>>>>  the offerer is willing to use when sending using the media, and 
>>>>>>> the
>>>>>>>  'humintlang-recv' values indicates the language(s) the offerer is
>>>>>>>  willing to use when receiving using the media.
>>>>>>>  -----------------new text---------------------------------
>>>>>>>  In an offer, the 'humintlang-send' values indicate the language(s)
>>>>>>>  the offerer is willing to select from for use when sending 
>>>>>>> using the
>>>>>>>  media, and the 'humintlang-recv' values indicate the 
>>>>>>> language(s) the
>>>>>>>  offerer is willing to receive one of in the media stream.
>>>>>>>  ----------------end of change----------------------------------
>>>>>>>  Motivation 1:) change from "indicates" to "indicate" in two 
>>>>>>> places to match the new use of plural "values".
>>>>>>>  Motivation 2:) Be sure to indicate that we only intend to 
>>>>>>> negotiate one language per media and direction, so that we do 
>>>>>>> not end up as unspecified regarding number of matches required 
>>>>>>> as the sdp "lang" attribute is.
>>>>>>>
>>>>>>
>>>>>>  Reworded.
>>>>>>
>>>>>>>
>>>>>>>  4. 5.2 Second paragraph
>>>>>>>  -----------------old text-----------------------
>>>>>>>  When a media is intended
>>>>>>>  for use in one direction only
>>>>>>>  ----------------new text---------------------
>>>>>>>  When a media is intended
>>>>>>>  for use for language communication in one direction only
>>>>>>>  ----------------end of change---------------------------
>>>>>>>  Motivation: Deletion of a note in this sentence made it less 
>>>>>>> obvious that we are only talking about directions of use of 
>>>>>>> language communication, and not about establishing asymmetric 
>>>>>>> media connections. Therefore add this clarification.
>>>>>>>
>>>>>>
>>>>>>  Reworded.
>>>>>>
>>>>>>>
>>>>>>>  5. 5.2 Deleted paragraph 6 before "Clients acting on behalf..."
>>>>>>>  ----------reinsert modified paragraph----------------------------
>>>>>>>  While signed language tags are used with a video stream to
>>>>>>>  indicate sign language, a spoken language tag for a video stream
>>>>>>>  indicates a request or offer to see the speaker, when that is of
>>>>>>>  importance for language perception.
>>>>>>>  -------------end of 
>>>>>>> change-------------------------------------------
>>>>>>>  Motivation: There was in the LC mail exchange a discussion 
>>>>>>> about sharpening up the specification of use of "unusual 
>>>>>>> combinations".
>>>>>>>  There was no agreement to delete them all. The one described in 
>>>>>>> this paragraph is the main one that has widespread use and needs 
>>>>>>> to be clearly specified for use by a large number of 
>>>>>>> hard-of-hearing and deaf users.
>>>>>>>
>>>>>>
>>>>>>  The text as it is now does not prohibit anything and explicitly 
>>>>>> mentions negotiating supplemental video by omitting language 
>>>>>> attributes on a video media.
>>>>>>
>>>>>>>
>>>>>>>  6. 5.2 Sixth paragraph
>>>>>>>  --------------------current text--------------------
>>>>>>>  (or for unidirectional streams, one of)
>>>>>>>  ------------------new text ------------------------
>>>>>>>  (or for asymmetrical use of languages, one of)
>>>>>>>  -----------------end of change----------------------
>>>>>>>  Motivation: We are not primarily talking about enabled 
>>>>>>> transmission directions of the streams, but about language use 
>>>>>>> in the streams. We do not want to limit the media stream 
>>>>>>> directions just because we do not specify an initial language to 
>>>>>>> use for that direction. There are other usage of media, and 
>>>>>>> there may even be occasional use of language in the direction, 
>>>>>>> just not worth mentioning as an initial and preferred use. The 
>>>>>>> suggested change should make that clear.
>>>>>>>
>>>>>>>  7. 5.3 Next to last paragraph
>>>>>>>  ------------------old text------------------------------
>>>>>>>  a list of supported languages.
>>>>>>>  -------------------new text-------------------------
>>>>>>>  a list of supported languages, media and directions.
>>>>>>>  -------------------end of change----------------
>>>>>>>  Motivation: It is not sufficient to know which languages are 
>>>>>>> supported, it is also essential to know in which media they are 
>>>>>>> supported and in which directions. (media could be replaced with 
>>>>>>> modality, but the media can become ambigous then, so use media 
>>>>>>> here to be brief.
>>>>>>>
>>>>>>
>>>>>>  I don't know that we can require this, but I'll add SHOULD kist 
>>>>>> supported languages and media. Demanding direction as well might 
>>>>>> be too unwieldy.
>>>>>>
>>>>>>>
>>>>>>>  8. 5.3, last line
>>>>>>>  --------------old text----------------------------------
>>>>>>>  Supported languages are: es, en"
>>>>>>>  --------------new text-------------------------------
>>>>>>>  Supported languages are: es, en transmission in audio; es, en 
>>>>>>> reception in audio"
>>>>>>>  ----------------------------------------------------------
>>>>>>>  Motivation: Same as for 7.
>>>>>>>
>>>>>>
>>>>>>  Fixed as above.
>>>>>>
>>>>>>>
>>>>>>>  9. 5.4 Undefined combinations
>>>>>>>  ----------------------------old 
>>>>>>> text--------------------------------------
>>>>>>>  The behavior when specifying a non-signed language tag for a video
>>>>>>>  media stream, or a signed language tag for an audio or text media
>>>>>>>  stream, is not defined.
>>>>>>>  ---------------------------new 
>>>>>>> text-----------------------------------------
>>>>>>>  There is no way specified for indicating use of text based 
>>>>>>> language in a video media stream.
>>>>>>>  There is no meaning assigned to specification of sign language 
>>>>>>> in an audio or text media stream.
>>>>>>>  --------------------------end of 
>>>>>>> change-------------------------------
>>>>>>>  Motivation: Seeing the speaker in video is an important 
>>>>>>> combination reinserted above in section 5.2.
>>>>>>>  This section therefore needed rewording to not include that 
>>>>>>> combination.
>>>>>>>
>>>>>>
>>>>>>  The draft explicitly mentions video for supplemental purposes.
>>>>>>
>>>>>>>
>>>>>>>
>>>>>>>  10. 6.2 Last sentence
>>>>>>>  -----------------current text---------------------
>>>>>>>  Supported languages are: [list of supported languages]."
>>>>>>>  -----------------new text------------------------
>>>>>>>  Supported languages and media and transmission directions 
>>>>>>> are:[list of supported languages and media and transmission 
>>>>>>> directions.]"
>>>>>>>  -----------------end of change--------------------------
>>>>>>>  Motivation: Same as for 7.
>>>>>>>
>>>>>>
>>>>>>  Fixed as above.
>>>>>>
>>>>>>>
>>>>>>>  11. 6.1 MUX Category
>>>>>>>  ----------old text in two locations-------------------
>>>>>>>  MUX Category: normal
>>>>>>>  ---------new text in same two locations--------------
>>>>>>>  Mux Category: NORMAL
>>>>>>>  ---------end of change-----------------
>>>>>>>  Motivation: Follow RFC 4566bis and IANA habits regarding use of 
>>>>>>> capitals
>>>>>>>
>>>>>>
>>>>>>  Fixed.
>>>>>>
>>>>>>>
>>>>>>>  12. 5.3
>>>>>>>  -------------old text-----------------
>>>>>>>  5.3 No Language in Common
>>>>>>>  -------------new text----------------
>>>>>>>  5.3 Preference parameter
>>>>>>>  ------------end of change 1 in 5.3---------------
>>>>>>>
>>>>>>
>>>>>>  The section is more than just the asterisk, it also advises use 
>>>>>> of specific SIP response codes if the call is failed.
>>>>>>
>>>>>>>
>>>>>>>  -------------old text-in 5.3, second 
>>>>>>> paragraph-------------------------------
>>>>>>>  The mechanism for indicating this preference is that, in an 
>>>>>>> offer, if
>>>>>>>  the last character of any of the 'humintlang-recv' or 'humintlang-
>>>>>>>  send' values is an asterisk, this indicates a request to not 
>>>>>>> fail the call.
>>>>>>>  --------------------------new text-------------------------------
>>>>>>>  The mechanism for indicating this preference is that, in an 
>>>>>>> offer, if
>>>>>>>  the last character of any of the 'humintlang-recv' or 'humintlang-
>>>>>>>  send' values is an asterisk, this indicates a request to not 
>>>>>>> fail the call.
>>>>>>>  The asterisk should be attached to attributes with languages of 
>>>>>>> lower
>>>>>>>  preference to be matched if such difference can be specified. 
>>>>>>> Thereby
>>>>>>>  the location of the asterisk can be used to support the 
>>>>>>> decision on
>>>>>>>  which languages to use in the call.
>>>>>>>  ---------------------------end of change 2 in 
>>>>>>> 5.3--------------------------------------
>>>>>>>  Motivation: There has not yet been any conclusion for my 
>>>>>>> proposal no 5 in the IETF LC comments of Feb 12.
>>>>>>>  This is a dramatically reduced version that may be easier to 
>>>>>>> accept at this stage, still covering one of the missing 
>>>>>>> functionalities in the draft.
>>>>>>>  The asterisk is used as a preference parameter in the 
>>>>>>> attributes. Thereby the proposed title change on 5.3
>>>>>>>  With this additional rule about where the asterisk(s) are 
>>>>>>> placed, the answering parties get good clues about the 
>>>>>>> preferences between alternatives presented by the offeror. The 
>>>>>>> chance to set up calls with satisfied users increase 
>>>>>>> dramatically compared to letting the answering party select by 
>>>>>>> chance between alternatives.
>>>>>>>
>>>>>>
>>>>>>  Making the asterisk a purely-advisory hint as to the 
>>>>>> least-preferred media/language combination seems harmless enough, 
>>>>>> as it would not be required to support it; however, I'm not sure 
>>>>>> it provides any benefit: if an offer contains some set of media 
>>>>>> with language, and the answerer can support all of them, should 
>>>>>> the answerer only include in its answer those without an 
>>>>>> asterisk? It seems simpler for the answerer to include everything 
>>>>>> in the offer that it can support.
>>>>>>
>>>>>
>>>>>  --
>>>>>  -----------------------------------------
>>>>>  Gunnar Hellström
>>>>>  Omnitor
>>>>>  <mailto:[e-mail]>[e-mail]
>>>>>  [phone]
>>>>>
>>>>
>>>>
>>>
>>
>>  --
>>  -----------------------------------------
>>  Gunnar Hellström
>>  Omnitor
>>  <mailto:[e-mail]>[e-mail]
>>  [phone]
>
>

-- 
-----------------------------------------
Gunnar Hellström
Omnitor
[e-mail]
[phone]