10 Mar 2011

See also: IRC log


Dan_Appelquist, Jeni_Tennison, Tim_BL, Noah_Mendelsohn, Jonathan_Rees, Ashok_Malhotra, Yves_Lafon, Larry_Masinter, Henry_Thompson


<scribe> scribe: Ashok

<scribe> scribenick: Ashok

Administrative items

<DKA> +lots welcome Jeni!

Noah: Welcome, Jeni

Jeni: Introduction

Noah: Change telcon timings?

Jeni: I think it will be ok

RESOLUTION: No change in telcon timings

Noah: Telcon on March 17 ... I am unavailable

<Larry> Leave meeting scheduled

<DKA> I am also available next week - and I think it would be useful given the number of things on our docket right now.

<Larry> If HT can, would go through IETF presentation

Jar: I can help with the agenda

<ht> I'm in favour -- hope to have IETF slide deck revised by then

Ashok: Regrets from me for 3/17

<JeniT> I can be here

<Larry> Just scheduling review IETF slide deck would be enough to meet

Noah: Jar will create agenda or cancel

Larry: We can review Henry's slide deck

Noah: Dan wanted to discuss f2f timing

Dan: There is a Social Web Summit Mtg in Berlin June 4/5. I'm very involved
... possible conflict ... could we meet a day later or the following week

Noah: Prefer not to change dates ... Dan may miss first day
... Tim, if you can meet a day later we can move it a day later

Dan: I can fly directly from Berlin to Boston

RESOLUTION: No change in June f2f dates

Approve Minutes from 3/3 http://www.w3.org/2001/tag/2011/03/03-minutes.html

RESOLUTION: Minutes from 3/3 are approved

TAG participation in IETF/IAB panel at March 2011 IETF (for March 2011 IETF meeting)

<noah> close ACTION-535

<trackbot> ACTION-535 Draft IAB meeting slide on scalability issues. closed

HT: I will create a new slide deck and we can discuss next week

<noah> close ACTION-517

<trackbot> ACTION-517 Figure out what to say about scalability of access at IETF panel closed

<noah> ACTION-530?

<trackbot> ACTION-530 -- Henry S. Thompson to draft slides for IETF meeting, with help from Larry Due 2011-02-22 -- due 2011-02-17 -- OPEN

<trackbot> http://www.w3.org/2001/tag/group/track/actions/530

<noah> ACTION-530 Due 2011-03-15

<trackbot> ACTION-530 Draft slides for IETF meeting, with help from Larry Due 2011-02-22 due date now 2011-03-15

<noah> ACTION-527?

<trackbot> ACTION-527 -- Noah Mendelsohn to make sure we make progress on ACTION-519 and ACTION-517 in time to provide input to Prague IETF meeting, talk to be ready by mid-March -- due 2011-02-17 -- PENDINGREVIEW

<trackbot> http://www.w3.org/2001/tag/group/track/actions/527

<noah> close ACTION-527

<trackbot> ACTION-527 Make sure we make progress on ACTION-519 and ACTION-517 in time to provide input to Prague IETF meeting, talk to be ready by mid-March closed

Larry: There have been discussions at IETF re. Registries etc. Brief mtg Sunday and breakfast mtg Thursday about Registries, Mime types, etc.

<ht> I will not be in Prague past Monday night, sorry

Larry: What does TAG want to happen wrt IANA registeries?

<Larry> it would be useful to have more of a mandate about what should happen with them

Noah: Perhaps add Registries to agenda for next week

ISSUE-60 (webApplicationState-60): Web Applications: Hash-bang (#!) URIs

Noah: Introduces the topic ... fuss about #! URIs ... some people feel that these have some downsides

<DKA> +1 to considering Jeni's post: http://www.jenitennison.com/blog/node/154

<noah> JT: People are using # URIs; a problem is that they are not sent to the server in cases where you might want them to be

Jeni: # URIs created a problem for Search Engines
... so they proposed a #! URI ... which takes the #! argument and converts to server parameters
... Twitter and Gawker use this
... problem was that JavaScript was not working
... problem also if Browser does not run JavaScript

<noah> JT: There are problems on various clients with JavaScript, including on my Galaxy Tab, some people don't have JavaScript, or turn it off. You don't get referrers, and you need Javascript to even see if the resource exists.

<timbl> I thought that the JavaScript happened to go down on the site was a bit of a red herring -- site can break in lots of ways -- but the issue of clients without JavaScript is a very valid one.

Jeni: Some architectural disadvantages to this pattern

<noah> Tim, I strongly agree that relying on JavaScript is a big issue, but FWIW Raman seems to disagree. In http://lists.w3.org/Archives/Public/www-tag/2011Mar/0096.html he says:

<noah> "at this point, JavaScript is part

<noah> of the Web platform, and trying to insist that URL references

<noah> make sense outside a javascript evaluation context does not make

<noah> sense on today's Web"

<noah> I find that quite unfortunate.

Jeni: We need to give some guidance

<Yves> Relying on JavaScript is as bad as relying on a specific codec for video, or image format, well, is it that bad? Depends on the intended audience

Noah: How good or bad is it to rely on JavaScript?

<Larry> JavaScript does not "go from a URI to a representation"

<DKA> Another datapoint: some webapp development tools, such as Google Web Toolkit, generate hash URIs as a default way to operate - this adds to the problem.

Noah: HTTP does not allow you to send frag id to server ... so no page reload if fragid changes

<Larry> The definition of HTML is changing to define the # fragment identifier **FOR HTML** to mean: for this content, send # parts to JavaScript.

<Larry> This doesn't have anything to do with HTTP actually

<ht> I am reminded of "a little bit pregnant" -- that the user experience depends on more than the material directly contained in the response message has been true for a long time

<ht> Maps? Really?

<Zakim> Noah, you wanted to talk about moving away from #

<Larry> We've put a lot of effort into an architecture which separates references, content types, and protocols, and am dismayed ...

<ht> +1 to TBL (based only on his q+)

Noah: We should promote HTML5 facilities that let you send URIs and do not cause page reload

Larry: We should not be sloppy about whether what we are discussing applies to HTTP, HTML or applies in general

<noah> Noah: my main point was that, where the item being referenced is a document-like (blog posting, twitter posting, map, etc.) we should assign them non-# URis as we always have. They should work consistently in Ajax and non-Ajax clients as necessary.

<Yves> Not really "sending to JavaScript" but fragment processing happens as described in the HTML spec and may also be interpreted by JavaScript present on that HTML page.

<jar_> Larry: don't want to lose that [...] orthogonality as we move forward on web architecture

<Yves> But clearly HTML's media type definition should be amended (and SVG's as well)

<Zakim> jar_, you wanted to say http: is not HTTP (Roy Fielding)

Larry: Consider HTML embedded in mail etc. where scripting may not be enabled

<noah> To me, having a URI that can only be dereferenced at the client and using JavaScript, for something as simple as a Tweet, seems very unfortunate.

<Larry> Noah, that may be unfortunate, but it's a site developer's choice, not an architectural lack

<noah> Larry, I think our role in the TAG is to advise those developers on the architectural consequences of the avaialble choices, and for my money, #! is really bad.

<Larry> Why is it unfortunate, Noah?

JAR: How the URI is processed is dependent on syntax

<jar_> Why does the URI pattern have to change when we move code from the server to the client or vice versa?

<Larry> Noah, i think we first need to articulate why it is 'bad', I'm having trouble coming up with reasons why it is bad

<Larry> So there's a picture on page 257 of a book, and all I have is the URI for the book, is it unfortunate that I have to use a fragment identifier to talk about that picture?

<noah> Seems to me it violates the rule of least power.

<Zakim> timbl, you wanted to say we are on the very of a possibly very complex (or satisfying?) architecture of things, pages, views, points of view, windows, queries, etc.

Ashok: JavaScript is now ubiquitous.

Tim: Is this the tip of an iceberg?

<JeniT> If JavaScript were actually ubiquitous, we wouldn't need the #! convention for search engines

<noah> Larry, it's unfortunate that if you were going to Amazon, the name of the book would be http://amazon.com/#book123

Ashok: Jeni, JavaScript in the client is ubiquitous

<JeniT> Ashok, aren't search engines clients?

<Larry> i'm thinking about the ISO 32000 use case (er, PDF), where you need uri-for-book.pdf#page=257 to access page 257

<noah> +1 search engines are absolutely clients with respect to the HTTP protocol that's used.

<Larry> These URIs are locators, not just identifiers

<Larry> HTTP URIs are locators

<noah> Larry, I don't think that example is the one that helps settle the case, because even with the traditional non-JavaScript Web you can argue it either way.

<jar_> TimBL: Use a language for expressing complex references, don't try to cram all expressiveness into URIs

<Larry> Noah, I'm not trying to settle the case, just discover some cases

Tim: #! works but it's a horrible kludge

<Zakim> timbl3, you wanted to mention (a) that the fragid is a really useful identifier and (b) that the JavaScript not being a first class language on the web is something which could be reengineered.

<Larry> Does it really have to be "JavaScript" or just "some scripting language"? Could #! work with Java?

<Larry> Agree with comments ... just hoping we can be more precise when we're talking about JavaScript vs. when we're talking about "active content" in general

Tim: Why did Google ask for a static piece why not use a Global Variable?

<noah> The TAG says [http://www.w3.org/2001/tag/doc/leastPower.html#plp]: Principle: Powerful languages inhibit information reuse.

<Yves> Active content that has access to the URI of the document

<Larry> Hypothesis: #! is a way of supplying some additional metadata that isn't otherwise obvious

<Larry> That the URI using #! is not just a locator but is intended as an indexible locator?

Tim: Making JavaScript a first class object would be a nifty idea!

Yves: The issue is that in the example Gawker they are not exposing a document but an application

<noah> YL: I think # per se isn't the point. What's happened is that sites like Twitter have gone from exposing linked documents, to exposing a single application that lets you see what would previously have been lots of documents

Yves: the shift is from exposing a document to exposing an application
... media types need to define fragid semantics

<noah> Strong +1 to what Yves says about the shift. That's a major loss for the Web, if those "documents" aren't properly linkable.

<Larry> Noah, so what is the problem with that? Yes, it's an application, no it's not a document. So?

<Zakim> DKA, you wanted to play the mobile card

Yves: Is #! available for all media types ... lot of things not nailed down ... need to be clarified

<Larry> They're linkable, just with #!... what's the problem?

<noah> Larry, just for a start, you can't crawl the links properly. Referrer doesn't work, right. See the whole list Jeni went over of what's broken.

<Yves> Problem is non-JavaScript aware clients (spiders, etc...)

<noah> DKA: JavaScript or not is not binary. Some device support it in a way you don't want to rely on heavily.

<JeniT> +1 or sometimes the Javascript support isn't complete

<Larry> Why can't referer (sic) be fixed, then? Isn't there some access to history ?

<noah> DKA: Lots of devices, agents, and other things the reference Web content that do not have JavaScript or don't behave as browsers do.

<Yves> Also it breaks in case of redirect

<noah> DKA: Includes things that produce thumbnails.

<Larry> Things that do not have JavaScript do not have HTML5, since HTML5 is an API to the DOM

<noah> DKA: Car apps that read you tweets

<Zakim> Noah, you wanted to remember rule of least poewr and to ask Tim a question and to remind rule of least power and ask Tim a question

<Larry> If you want thumbnails or to index HTML you need a javascript interpreter

<DKA> I like JavaScript just fine, by the way. Some of my best friends are JavaScript developers.

Noah: Using JavaScript all the time violates rule of least power

<Larry> Yes, there's a tradeoff when you design your site about whether you want to make people use JavaScript.

Noah: JavaScript is a powerful language and perhaps too powerful
... asks Tim "Should we says JavaScript always is there and use it and if it's not there too bad"?

Tim: The rule that you don't reload page ... may go away

<jar_> A problem is inferring operational behavior from the syntax of the URI. This inhibits URI reuse, since URIs have to change when deployment details change. if method for 'dereference' were decoupled from URI syntax, URIs would be cooler. Routing problem.

<noah> Noah: what I said was, we have published the Rule of Least Power, which says effectively "don't use Javascript, even if it's there, if HTML will do"

<JeniT> http://www.w3.org/TR/html5/history.html#the-history-interface

Tim: When you select an item on the screen or highlight it, it goes on the address bar ... items may be from different domains

<JeniT> It has support in Webkit as well as Firefox I believe

Tim: Are there any constraints when you do a pushstate()?

<Zakim> Larry, You wanted to talk about architecture that is less judgemental ("#! is bad") and more clearly defining consequences

Larry: We will not get a #! is good/bad finding ... we can discuss the consequences

<noah> We absolutely can say #! is bad in the following ways, and either "don't do it" or "don't do it unless..." Web Arch says things like that all the time.

Larry: #! adds some metadata wanting fragid to be converted to server-side params

<ht> I don't agree

Larry discusses use case where you evolve the app using #

<Zakim> ht, you wanted to enlarge the issue

Noah: The AJAX app would need to create a URI for example for the email you selected

HT: Focussing on page refresh is too narrow ... there are larger class of issues hidden behind this ... interaction with search engines
... AJAX gave good interactivity but you lost indexing

<noah> Henry, I hear you, but I don't see how that explains why they went first to # (which I claim is to minimize server interactions) and then #! (which was to get back indexing, given that client limitations forced them to use #)

HT: We need to think about the various players in the Web world
... lets brainstorm at next f2f

<Zakim> JeniT, you wanted to mention distributed applications

HT: I will try and send mail

Jeni: Sometimes the data comes from more than one server

Yves: An application is a portal that gives you access to information ... it's not one server

<noah> TBL: I find it frustrating that with Twitter, you're interesting in 140 characters, you have framesets with content coming from another document. You want the inner frame to be associated with what's in the address bar.

<noah> TBL: I'd like to see someone write the alternative version where you get the Tweet as you want it more or less pure, then the javascript loads and surrounds it.

<noah> TBL: In the case Jeni mentions, where an app composes content from several sources, we should push for interoperability not so much at the [id of the state of the app???] level, but at the data interop level.

Tim: When data comes from several sources we would focus on data interoperability

<Zakim> Noah, you wanted to propose next steps for Ashok

Noah: Next steps on HashInURI
... several points of view
... why not summarize different positions and their pros and cons

<DKA> +1 to merging in some of Jeni's post.

<timbl> +1

<Larry> Definitely want to talk about registries, and IETF presentation on webarch for webapps

Summary of Action Items

[End of minutes]

Minutes formatted by David Booth's scribe.perl version 1.133 (CVS log)
$Date: 2011/03/23 15:13:31 $