Monday, 5 November 2012< ^ >
Scott Hollenbeck has set the subject to: WEIRDS WG at IETF-84
Room Configuration

[21:48:44] Dowon Kim joins the room
[21:53:36] Dowon Kim has set the subject to: WEIRDS WG at IETF-85
[22:04:44] SM joins the room
[22:20:08] Dowon Kim leaves the room
[22:23:28] danyork joins the room
[22:24:24] Guangqing Deng joins the room
[22:25:52] john.levine joins the room
[22:25:59] john.levine leaves the room
[22:29:21] joseph.yee joins the room
[22:31:50] Linlin Zhou joins the room
[22:33:41] Dowon Kim joins the room
[22:34:18] naptee joins the room
[22:34:46] naptee leaves the room
[22:35:41] naptee joins the room
[22:40:50] dseomn joins the room
[22:41:18] =JeffH joins the room
[22:41:53] fneves joins the room
[22:42:08] Andrew Sullivan joins the room
[22:42:10] Francisco Arias joins the room
[22:42:11] john.levine joins the room
[22:42:14] jpdionne joins the room
[22:42:14] yone joins the room
[22:42:17] Arturo Servin Ü joins the room
[22:42:20] <Andrew Sullivan> Hi, I'll do the jabber scribe
[22:43:02] <Andrew Sullivan> Note well
[22:44:01] Suz joins the room
[22:44:11] <Andrew Sullivan> Update from ICANN
[22:45:05] <Andrew Sullivan> Francisco Arias at the mic talking about what happened at ICANN in Toronto with respect to WHOIS and WEIRDS
[22:45:12] Benno Overeinder joins the room
[22:46:01] AK joins the room
[22:46:26] <Andrew Sullivan> Andy Newton at mic about why ccTLDs ought to implement
[22:46:44] <Andrew Sullivan> Peter Koch at mic
[22:46:54] joins the room
[22:47:10] <Andrew Sullivan> asking not to rely on "regulator"
[22:47:29] <Andrew Sullivan> Olaf Kolkman suggesting that the best protocol is the selling point
[22:47:39] <Andrew Sullivan> John Klensin making point about this at mic
[22:48:06] <Andrew Sullivan> "be careful"
[22:48:58] <Andrew Sullivan> Barbara Roseman @ mic. Only some ccTLDs under regulation.
[22:49:06] <Andrew Sullivan> all moving towards same goal: replace WHOIS
[22:49:24] <Andrew Sullivan> Mics closing. Klensin at mic: Barbara part of IANA? No
[22:49:30] <SM> :-)
[22:49:31] byron.ellacott joins the room
[22:49:32] <Francisco Arias> only some ccTLDs under contract, that's what Barbara said
[22:49:52] <Francisco Arias> under contract with ICANN
[22:49:53] <Andrew Sullivan> Frederico Neves at mic: please move on. ccTLDs in room
[22:50:03] <Andrew Sullivan> sorry, yes, "contract" not "regulation"
[22:50:19] JimG joins the room
[22:51:07] <Andrew Sullivan> Turning to JSON response now
[22:51:50] <Andrew Sullivan> Nobody sees anything missing, turning to format &c.
[22:51:58] <Andrew Sullivan> John Levine going to mic
[22:52:48] <Andrew Sullivan> parts of current draft not clear: what mandatory and not?
[22:53:39] <Andrew Sullivan> Olaf: from last meeting, everything is mandatory to implement but not serve
[22:53:52] <Andrew Sullivan> Levine: interoperability problem
[22:54:27] <Andrew Sullivan> Hum: who is prepared?
[22:54:52] <Andrew Sullivan> 3 people feel the draft has all the elements (I think that's what the question was)
[22:55:01] <Andrew Sullivan> chair exhorts to read document and send comments
[22:55:12] <Andrew Sullivan> [name?] support what said
[22:55:20] <byron.ellacott> I think audio is lagging reality by more than a few seconds
[22:55:23] <byron.ellacott> arturo servin at the mic, btw
[22:55:36] <Andrew Sullivan> comfortable with document in numbers world, not names
[22:55:38] <SM> ariubd 20 seconds
[22:55:51] <Andrew Sullivan> Jim Galvin at mic: nothing obviously missing
[22:56:04] <Andrew Sullivan> but maybe things missing will be discovered later during implementation
[22:57:39] Peter Koch joins the room
[22:58:47] <Andrew Sullivan> Scott Hollenbeck at mic. Verisign implementing from server side
[22:59:07] <Andrew Sullivan> thin registry — need stuff from thick model
[22:59:46] <Andrew Sullivan> procedural questions: not chartered to produce data model, so might send through ISE
[22:59:57] <Andrew Sullivan> Pete Resnick, does not remember conversation
[23:00:12] JcK joins the room
[23:00:19] <Andrew Sullivan> discussion about charter
[23:00:32] <Andrew Sullivan> can inventory come out of WG?
[23:00:37] joins the room
[23:01:02] <Andrew Sullivan> shouldn't be that controversial
[23:01:50] <Andrew Sullivan> Next, WEIRDS response could always be empty
[23:02:01] <Andrew Sullivan> in which case WEIRDS isn't much better than WHOIS
[23:02:09] <Andrew Sullivan> and need some sort of applicability statement?
[23:02:17] <byron.ellacott> if you narrowly define WHOIS's problems as "could produce an empty result"
[23:02:25] <Andrew Sullivan> Resnick: carve carefully
[23:02:33] <Andrew Sullivan> avoid policy
[23:02:44] <Andrew Sullivan> implementation guidance only
[23:03:10] <Andrew Sullivan> Next, Andy Newton on Internationalization
[23:03:25] <Andrew Sullivan> I18N in RDAP slides
[23:03:47] <Andrew Sullivan> nobody is an internationalization expert
[23:04:04] <Andrew Sullivan> Slides: Character Sets
[23:04:16] <Andrew Sullivan> 9if there are slide numbers, I can't see them)
[23:04:28] <Andrew Sullivan> Slide: Language Designation
[23:06:19] <Andrew Sullivan> Slide: Desired Language
[23:06:50] <Andrew Sullivan> Slide: Script & Charset Designation
[23:07:44] <Andrew Sullivan> Slide: IDNs (appendix C. . .)
[23:08:12] tony.l.hansen joins the room
[23:08:38] <Andrew Sullivan> (I note that John Klensin at the mic taking notes.)
[23:08:51] <Andrew Sullivan> Slide: Postal Addresses
[23:09:34] <Andrew Sullivan> Slide: please help
[23:09:39] <Andrew Sullivan> John Klensin at mic
[23:10:03] Arturo Servin Ü leaves the room
[23:10:27] <Andrew Sullivan> Slide: Language Designation
[23:10:44] <Andrew Sullivan> jck talking about degrees of precision in language
[23:11:30] Arturo Servin Ü joins the room
[23:11:56] <Andrew Sullivan> language identification standards problematic
[23:12:30] <Andrew Sullivan> Pete Resnick at mic
[23:12:53] <> One of the use cases for indicating language/script of the contact information is for the internationalized contact information, it allows people who do not speak the language or know the script to be able to transliterate or translate.
[23:13:12] <Andrew Sullivan> steve: for mic?
[23:13:37] <Andrew Sullivan> jck again: that's harmless if it's just markup
[23:13:44] <Andrew Sullivan> slide: desired language
[23:13:54] <Andrew Sullivan> Slide: script & character set designation
[23:14:12] <Andrew Sullivan> specifying UTF-8 is painful
[23:14:56] vincent.levigneron joins the room
[23:15:06] <Andrew Sullivan> (@steve.sheng: I assume you didn't say "for mic" on purpose, so I won't go there unless you say.)
[23:15:58] <byron.ellacott> I think that information was conveyed by Andy anyway
[23:16:00] <Andrew Sullivan> Pete Resnick again
[23:16:21] <Andrew Sullivan> are there registries that want to serve data & are going to want to serve data in that way?
[23:16:40] <Andrew Sullivan> Andy: historically yes, but might be willing?
[23:17:09] <byron.ellacott> some picked those character sets because they were a reasonable choice at the time, when UTF-8 wasn't particularly widespread; probably not important for the mic
[23:17:25] <Andrew Sullivan> Klensin again
[23:17:34] <Andrew Sullivan> script issue: some languages written in more than one script
[23:17:52] <Andrew Sullivan> for readers, mostly can read both scripts, have rendering devices, &c.
[23:18:19] naptee leaves the room
[23:18:33] <Andrew Sullivan> [I don't understand why this matters if we're going to use IETF language tags, though, since they can have script subtags. But I'm not going to the mic.]
[23:18:42] naptee joins the room
[23:19:01] ray.bellis joins the room
[23:19:13] <ray.bellis> Isn't UTF-8 actually mandatory for JSON ?
[23:19:18] <byron.ellacott> no
[23:19:26] <byron.ellacott> unicode is mandatory for JSON, but it can be -8, -16, or -32
[23:19:56] <Suz> Olaf: is there a way we can produce a spec that is forward compatible with solving i18n issues?
[23:20:13] naptee leaves the room
[23:20:14] <Suz> clarification requested, "can we write a spec that defers these problems to later?"
[23:20:25] <ray.bellis> sorry, I was imprecisely remembering RFC 4627. It appears to mandate Unicode, but in any of several common encodings
[23:20:29] <Suz> Klensin: analogous to retrofitting security
[23:20:31] <> RFC 5646 can probably satisfy the tagging requirements.
[23:20:50] AK leaves the room
[23:21:06] <Suz> Andy Newton: is the script issue a solved problem?
[23:21:17] naptee joins the room
[23:21:21] <byron.ellacott> I would think 5646 *should* cover it
[23:21:22] <Suz> Klensin: tower of babel ref
[23:21:49] <Suz> (guest scribe while Andrew is at the mic)
[23:22:10] <Suz> ??: have you looked at vcard? new RFC on vcard 4, thought about all of these
[23:22:11] naptee leaves the room
[23:22:29] naptee joins the room
[23:22:29] <Suz> some of the vcard work could be re-used
[23:22:42] <Suz> (Peter Saint Andre draft, url anyone?)
[23:23:01] <Suz> data model is rasonably mature on i18n
[23:23:32] <Suz> Peter Koch: what language is my name in? as illustration of the meta-problem
[23:24:05] <Suz> Peter: 2 issues: representation, but also semantics of names
[23:24:06] AK joins the room
[23:25:30] <Suz> Peter: all kinds of engienering/operational issues with context
[23:27:04] <byron.ellacott> the semantics of the HTTP header field is for the preferred/intended *consumption* language, not content language
[23:27:43] <Francisco Arias> @Byron, what is the difference between "consumption" and "content" language?
[23:27:46] <Suz> Pete Resnick, no hat: distinction between what's returned i nthe data and what the client is asking for; ? semantics about query/response as separate from data, which does have a language tag (thanks Byron, this is subtle stuff)-- this is exactly the distinction Pete is making
[23:28:06] <byron.ellacott> farias, the usual example is that a "learning Japanese" page for English speakers will have Japanese content, but be intended for an English speaker
[23:28:45] <ray.bellis> @farias it appears to be a question of granularity
[23:29:13] <Suz> Andy: no we don't make this distinction, between data we expect to be translated and any data that is being presented in its language of origin
[23:29:33] <Suz> Olaf: almost implies that for query you don't need this
[23:29:48] <Suz> Pete: this may be true, if all you're getting b ack is data not intended for translation
[23:31:02] <Suz> Andrew Sullivan: slightly worried we're being overly alarmist here; not underappreciating the complexity, but we're talking to repositories that have rules about what's in there and clients that act within constraints
[23:31:47] <Suz> Andrew: to answer the question about deferring the complexity is "well, sort of", but tagging everything undefined but still collecting the relevant data that may be usable later.
[23:32:02] <Suz> Not creating a new problem
[23:32:31] <Andrew Sullivan> Joseph Yee: why are we worrying about translation and transliteration?
[23:32:36] <Andrew Sullivan> there's no guarantee they're right
[23:32:54] <byron.ellacott> [notformic] for me, I'm not worrying about translation/transliteration at all :-)
[23:33:03] <Andrew Sullivan> on future compatibility: how far forward are we looking?
[23:33:20] naptee leaves the room
[23:33:42] <Andrew Sullivan> John Klensin again
[23:34:04] <Andrew Sullivan> trying to explain how bad things are
[23:35:22] naptee joins the room
[23:36:56] naptee leaves the room
[23:37:03] <ray.bellis> I vote for "name: <chinese version> / <some other version>"
[23:37:17] <ray.bellis> and no need for all the rest of these complications
[23:37:25] <Suz> (guest scribe is back)
[23:38:16] <Suz> Olaf: isn't the complexity about what to answer to a query a matter of registry policy?
[23:38:38] <Suz> Klensin: yes but you have to decide which 2 or 3 or more to store (transliteration, original, translation, ....)
[23:39:19] <> translation and transliteration is about the data, is out of the scope of the WG, right?
[23:39:23] <Suz> Klensin: this is about names, phone numbers may also be complex
[23:39:28] <byron.ellacott> steve, specifying a need for it is
[23:39:43] <byron.ellacott> steve, but some registries already have transliterated/translated data stored, and the protocol should probably be able to handle them
[23:39:45] <Suz> Judith Vasquez: perspective on user case
[23:39:57] <Suz> law enforcement and intellectual property
[23:40:27] <Suz> thus language of origin is most critical
[23:40:57] <ray.bellis> origin being the one supplied by the registrant?
[23:41:00] <Suz> Olaf, no hat, responding: for particular constituencies this is a key use case, but not the only key ones
[23:41:04] <Suz> Judith: agreed
[23:41:07] joseph.yee leaves the room
[23:42:29] <Suz> andrew Sullivan: frustrated with the way we're exploring potential disasters; of course they're potentially infinite, but the repository has only so much data that can come out in scary ways
[23:42:43] <Suz> Olaf: what goes into the db is policy
[23:43:01] <> Judith is correct, if we don't get this right, there is no hope for this protocol to be adopted at ICANN.
[23:43:03] <Suz> Andrew: yes and we agreed we enable but not make policy. Limited set of operations required
[23:43:13] <Suz> Steve: her concern is policy tho
[23:43:24] <Suz> (sorry, that was me to @steve.sheng)
[23:43:35] <> @suze, yes, agree.
[23:43:46] <byron.ellacott> suzanne, so the protocol should *enable* that policy decision, right?
[23:44:35] <ray.bellis> get everything back, use Google translate on stuff you can't parse...
[23:44:47] <byron.ellacott> which, if I understood Judith's point, means the protocol should be able to tag the language of the response (or part thereof) as specified by the creator of the data, which may be a registrant rather than the registry itself - but I have no beer, and may not be following well
[23:44:49] leaves the room
[23:45:11] <SM> She said that it is important to have the original stuff
[23:45:19] <SM> fro LEO and the otehr crowd
[23:45:52] <byron.ellacott> oh right, the protocol shouldn't force a transformation of the data, I can totally agree with that
[23:46:11] <Andrew Sullivan> Peter Koch: if there needs to be a selection mechanism, ought to be in first version of the protocol
[23:46:22] <SM> Hmm, she was looking at this froma different angle, if I understood:)
[23:46:23] <Linlin Zhou> display the information according to users’ original input, and could also have translation or transliteration
[23:46:36] <ray.bellis> @bryon script transformation, or chatset ? ;-)
[23:46:38] <ray.bellis> charset
[23:46:39] naptee joins the room
[23:46:45] <SM> stop mentioning charset
[23:46:46] <byron.ellacott> ray, semantic transformation
[23:46:52] <byron.ellacott> charset is not semanti
[23:46:55] <Francisco Arias> @Byron, translation or transliteration of the data is outside the protocol, agree
[23:46:55] <Andrew Sullivan> Peter: back to original question: preference or requirement from query side
[23:46:56] <SM> It is bad karma during this part of the discussion:)
[23:47:01] <Andrew Sullivan> not addressed completely
[23:47:19] <SM> so what Pete suggested is not gaining traction?
[23:48:29] <Andrew Sullivan> We're back into transliteration again
[23:48:30] joins the room
[23:48:35] <byron.ellacott> my understanding of what would be useful here, is that if the user asks for language/script xxXX, and there are two variants *stored* as provided by the origin, and one happens to be xxXX, provide that variant, else fall back to your default
[23:48:37] <Andrew Sullivan> I don't understand how we got here
[23:48:54] <SM> Yes, but then the side discussion started
[23:49:00] <SM> Not sure where we are now
[23:49:24] <Andrew Sullivan> @byron.ellacott: of all the words to pick, "variants" is probably the most pregnant.
[23:49:35] <byron.ellacott> I'm a numbers person :-)
[23:49:55] <SM> Wait, your region has scripts:)
[23:50:13] <> what language or script the registration data is in is a registry choice, it is not a protocol issue.
[23:50:14] <ray.bellis> @byron I agree with your understanding - give the preferred language if you have it, else return your local default
[23:50:15] <Andrew Sullivan> Pete Resnick at mic
[23:50:24] <Andrew Sullivan> trying to sort out issue
[23:50:36] <Andrew Sullivan> "server would transliterate on fly". Is that a requirement?
[23:50:40] <Francisco Arias> or return everything and let the application present what better suits the need of the client
[23:50:42] <Andrew Sullivan> No.
[23:50:51] <> No
[23:50:54] <SM>
[23:51:00] <ray.bellis> Hell no!
[23:51:00] <byron.ellacott> steve, when the registration data is in multiple languages/scripts, it is a protocol issue to select which one(s) you want, and inform the client of what you're carrying, I think
[23:51:04] <Francisco Arias> no, transalation/transliteraion should be outside protocol
[23:51:06] <Andrew Sullivan> Say "can't request particular language"
[23:51:12] <Andrew Sullivan> and return anything you want
[23:51:26] <Andrew Sullivan> client's responsibility of what to do
[23:51:28] <Andrew Sullivan> but each marked
[23:51:29] <> @byron, but that's a client responsiblity, right?
[23:51:51] <ray.bellis> We only support English - why would we mark each individual element?
[23:51:56] <byron.ellacott> steve, one way to solve the problem is to send them all and let the client pick what to display
[23:52:09] <byron.ellacott> steve, another way to solve the problem is to send only one but indicate others are available
[23:52:19] <Andrew Sullivan> are there registries that provide hundreds of different rep'ns of every object?
[23:52:28] <Andrew Sullivan> Klensin again
[23:52:48] <Andrew Sullivan> @ray.bellis: so have a hierarchy of tags
[23:52:52] <Andrew Sullivan> so you can inherit from the top level tag
[23:53:21] <Andrew Sullivan> Klensin: non-issue of on the fly generation
[23:53:46] <Andrew Sullivan> Klensin: some advocates want extravagant data requirements
[23:55:15] <Andrew Sullivan> Protocol design issue:
[23:55:34] <Andrew Sullivan> could be registries that insist on returning large number of different versions of data
[23:55:50] <Andrew Sullivan> ok w/ WG will create very poor experience?
[23:56:24] <Andrew Sullivan> (asking as AD)
[23:56:34] <Andrew Sullivan> Sorry, didn't id speaker: Pete Resnick
[23:57:41] <Andrew Sullivan> Klensin: describing protocol where someone can ask a question with unspecified id, where reg response with what it feels like: how is this better than Whois?
[23:58:11] <Andrew Sullivan> is it valuable enough?
[23:58:46] <Andrew Sullivan> Pete again, AD comment
[23:59:25] <Andrew Sullivan> Charter is clear on what the work problem, & John's question is "worth doing?", and AD says it sounds like a meta discussion, not a WG work item
[23:59:28] <> @byron, yes, I see your point now, thinking.
Powered by ejabberd Powered by Erlang Valid XHTML 1.0 Transitional Valid CSS!