5 Jul 2005

See also: IRC log


Henry S Thompson, Vincent Quint, Dan Connolly, Tim Berners-Lee, David Orchard (in part), Dean Jackson (in part)
Roy Fielding, Noah Mendelsohn, Norm Walsh, David Orchard(partial), Ed Rice
Vincent Quint
Henry S Thompson



<DanC> expecting Ed and Timbl

DO to scribe next week

<DanC> regrets NM for 12 July


TimBL: Existing issue XMLProfiles-29. . .

<timbl> http://lists.w3.org/Archives/Public/www-tag/2002Nov/0156

<DanC> Issue withdrawn by the tag http://www.w3.org/2001/tag/2004/11/29-30-tag#item17

TimBL: we withdrew our request to XML Core to do something
... but not clear that the issue is really dead

<DanC> (the norm is that the TAG announces a decision to close an issue, saying to the originator, "ok with you?")

TimBL: need to check with Paul Grosso and Norm Walsh

<timbl> There should be a message to say what the tag resolved.

<timbl> Or that it was withdrawn by Paul G

<scribe> ACTION: Chair to check status of XMLProfiles-29 with Paul Grosso [recorded in http://www.w3.org/2005/07/05-tagmem-minutes.html#action02]

Web Applications

VQ: Thanks to Dean for joining us so late in the day in his timezone
... [background to TAG interest in Web Applications]

<dino> http://www.w3.org/2004/07/webapps/webapps

DJ: Draft charter is in W3M review
... WebApps have become cool again in the last 12--6 months, particularly AJAX, based on XMLHttpRequest

<timbl> HTML, DOm, Javascript, XMLHttpRequest

DJ: Technology has been around, but only crystalised recently
... Gmail is a good example of HTML+DOM+Javascript
... Whole mailbox is one page, mail comes in/out via Javascript plus HTTP requests behind the scenes
... Same URL throughout the whole session
... Another example is Google Maps
... Dashboard widgets. . .
... Devs complaining about limitations, requests coming in to W3C for support/features
... Some of this has filtered in to our specs, e.g. SVG

<DanC> hmm... yes... what deliverables/scope?

DJ: But SVG is not really the right place for a Network API -- we maybe need a REC which addresses these issues across all the relevant domains

TBL: Is this an ECMAScript-based story or not
... Scope says just "client-side applications for the web" -- that covers Perl, Haystack etc.
... But I thought this was about refining/standardising existing hacks, i.e. standardising Javascript usage of today

<timbl> Dean: The DOM is used in Javascript, but in Python one could do a much better specific language binding.

DJ: Yes. C.f. the reaction of the dev community to the DOM -- we're only going to use it through Javascript, so the cross-language stuff is a pain

<timbl> It should be extending and refining existing practice ... but Java is often used when Javascript is not on phones.

DJ: So mostly yes, Javascript, but phones are a worry, because Java, not Javascript, is the tool of choice there
... Also, consider XMLHttpRequest - not needed in Java, it's already there
... Cookie-support OTOH is something which makes sense in the browser context, not in the general web/pgming language interface
... The focus is on standardising across the current diversity of Javascript usage into specific W3C-blessed APIs

<DanC> (hmm... in fact, skip the specs and go straight to the test suite. just give this WG the mandate to decide the outcome of test cases presented by the community. 1/2 ;-)

TBL: The charter should say this -- no sense beating around the bush, stop people from joining to design a new language in their tracks

DJ: Take that on board

TBL: First deliverable is a list of deliverables?

DJ: That came from W3M input

HST notes that writing a Requirements doc't is often the way a new WG starts. . .

TBL: I'd worry that that opens the door to new languages
... Compare what's happened to CDF -- embedding is the interesting and important bit, but they're focussing on 'by reference', because it's the easy part

DJ: Well, we hope the problems we're solving wrt 'by reference' will make the embedding part easier when we get to it

TBL: Scoping is really important. . .

DJ: Can easily make it less Javascript specific

TBL: But is that really the right direction to go -- don't you lose fluidity if you move away from Javascript-specific -- consider E4X, tightly coupled to Javascript and very clean as a result

DJ: Compare Java, strong-typing, vs. Javascript and Python, can change type at runtime
... Main usecase is Javascript, I'm happy to put that in the charter

<Zakim> DanC, you wanted to ask about how many programming languages in the web...

TBL: People have said the the DOM is clumsy -- OK for all languages, but good for none.

DC: E4X looks good, cool if runtime understands it, but it's my understanding that the huge bulk of Javascript art is figuring out what support you've got for what features and how you can do business
... Architectural question: How many programming languages for the web -- Ref. Programming Languages workshop in 1995 (ref below), concluded in favour of 'many'. But by 1997, with COM, CORBA and Java in play, with MSoft opposed to anything beginning J*, we then went on to do our first, namely XSLT
... So 'how many' is still an interesting question

<DanC> Mobile Code Workshop: 05 July 1995

DJ: Answer: 1 -- even phone platforms are moving towards Javascript
... [something] from Microsoft coming along for phones, embed any .NET language

DC: Languages I run in to: Flash/Lingo

DJ: Lingo is changing to be more like Javascript
... Nice pgrming environment, simpler object setup than HTML

<DanC> ActionScript is more like...

DC: ActionScript is compiled out, there's XSLT, Java . . . not growing all that fast in my experience
... Biggest difference between Javascript and Python is that Javascript shares code by copy and paste, because of security restrictions

DJ: But the <script> tag can reference anything and it will get downloaded
... But once you're running, the code can only retrieve from the place the matrix page came from
... Reason is to stop people snarfing stuff from inside the firewall once they get there

<Zakim> DanC, you wanted to ask about javascript importing: namespaces and security

DC: Is Javascript's namespace flat?

DJ: Yes -- the charter talks about the Window object, which is not standardised anywhere but implemented in all browsers
... If you expect to share code, use xx_ in front of all your names

TBL: Is the Window thing in the charter?

DJ: Yes, called Global Browser Object

TBL: Any pointer to anyone's documentation of this would be good. . .

DJ: Yes, 2nd deliverable is a "State of the WebApp World" -- that could include e.g. a Wiki collected by Team cataloguing everything we can find . . .

<DanC> hmm... TAG finding ?

DJ: It would be good IMO to move this off the WG and onto the Team
... Question: Google Accelerator pre-fetched links from search result, but this caused trouble by stimulating lots of server-side code
... Non-bookmarkable stuff that just blew up

<DanC> (er... GET vs POST is one issue (#7) and bookmarkable-cookie-state is another (not on our list yet?))

TBL: The principle of GA is OK

[HST doesn't understand why POST would be better]

DJ: Lazy use of GET to save 2 lines of code

<DanC> (using POST in the right places isn't helped by some aspects of HTML)

VQ: When we discussed this last week, NM asked about open architectures wrt WebApps -- comment?

DJ: Not sure what he meant by that

DC: Worried about monopoly control of the design space

DJ: Everyone shares that concern: An open-standard document format that can do text animation graphics etc.
... And an open-standard programming environment
... Javascript is good enough, maybe XUL is good enough, but nothing comes close in terms of current usage to HTML+Javascript

VQ: NM also concerned about client-side memory requirements

<DanC> (dean, if it's been in IE for 7 years, surely there's some docs somewhere)

DJ: Think that's a misunderstanding, hope current charter clarifies, persistent storage is not a requirement

<dino> there are. somewhere on msdn.com

<dino> I'll look

TBL: Make sure there is more than one line in the charter in terms of answering this sort of question, so that scope is clear

DJ: RF mentioned security, something will go in charter

<Zakim> ht, you wanted to ask about POST

<DanC> (using GET for "buy this album" is lazy)

HST: Why is GET the wrong thing?

DJ: Use <a> in a ToDo list as a way to build a 'done' button
... So prefetching from the todo list dropped all the items

<Zakim> timbl, you wanted to ask about Dean's personal prioritization of these issues, and interdependencies

TBL: Suppose you were chair, how do you proceed? Fork multiple taskforces, because things are independent?

DJ: Yes, multiple task forces would be the plan, group as a whole would not have much to do
... Most important bit is XMLHttpRequest -- WHATWG is also doing this, as part of HTML5, we could/should do it just on its own
... Another important bit is DOM level 3 Events, it's in Process limbo, needed in lots of places
... Lots of interest in XML UI, I'm not really up on that, lots of folk are concerned to get a standard before we get a lot of non-standard ones

TBL: Start from XUL?

DJ: Maybe, I don't really know enough

TBL: Separate WG?

DJ: Maybe -- there is a risk the WebApp WG gets all the bits that have no home
... I'd rather make this a small group with small target and get it done
... So keep XML UI out of it

TBL: [missed this]

DJ: Let's see what the AC review says

DC: There's a risk, at formal review stage, that they just say 'yes' and you're stuck with it all

<DanC> TBL: to give the same group both (1) a list of small manageable tasks and (2) a big open-ended thing is to put both at risk

TBL: This all sounds right, you've got the energy pointing in the right direction, XUL is the outsider, best to leave it out

DJ: Right, don't want to get bogged down -- focus on refine and standardise existing practice in the API area/runtime library

TBL: But not a new declarative language for UI, i.e. XUL

DJ: Another relevant point is XBL2, could connect up with taskforce in WebApp WG

DJ: Implemented in Mozilla

<DanC> XBL (Extensible Binding Language) 1.0

DJ: So I hear [TBL] with his W3M hat on suggesting moving the XML UI stuff out of this charter

TBL: Yes, needs more investigation and isn't as ready yet

<DanC> note also... SVG's XML Binding Language (sXBL)

DO: Is managing the growth and manipulation of complex state in Javascript to protect against the loss of URIs for the results on the table?

DJ: No, that's not on anyone's radar

DO: So the long-term effect of these problems is not an issue

DJ: Similar to Google Accelerator problem, they did a simple obvious thing but it had unexpected negative consequences

DO: Is there anything the TAG can do about it? How can we help make there be more URIs/make the Back button work

DJ: People are using Ajax because they believe the user experience is enhanced because it makes things so much faster than reloading the whole page
... That view is in direct tension with making the Back button work

DC: These guys are savvy wrt what they're doing, but consider when that approach/toolkit is used for a bank website and we move from my statement to a deposit etc. w/o a change in URL

DJ: Well, guidelines on what you shouldn't do would be a good idea -- for example "Don't use GET for things that change the world"

<DanC> (for webaps... forget specs... build a cool "planet webapps", with a validator, complete with a button "I'm surprised by the validator results, and I hereby donate this to w3c as a test case")

TBL: Question for the TAG: Are we happy with the Web losing its declarative style, more and more going in to Javascript?

DC: No not happy, see Tim's pending action wrt Principle of Least Power

DJ: There are some really bright Javascript hackers who, e.g. to populate a page, will use microparsing of info in the page and templates to do the job
... So even w/o scripting you can get some sense of what the page is
... WHATWG is making very good progress on that front
... TV Raman said, W3C's job is to watch what's happening on the Web, take non-declarative usages and move them into a declarative form -- XForms is his positive example here

VQ: So will WebApp WG feed back to XForms?

DJ: Not as such, but I do want to know what the XForms people think about WebApps

TBL: Could you write XForms using WebApps?

<DanC> (there's some company that has done an XForms implementation in javascript... what's it called?)

DJ: Sure, mostly if not all -- Schema?

TBL: So is that a kind of transition strategy?

DJ: Yes, try to put as much declarative stuff into the doc't as possible, a bit of script to do the clever stuff

HST thinks Dave Ragett's SLIDY is a similar story

<dino> yes

HST: If WHATWG are targetting the 80/20 point and doing it declaratively, why aren't we?

DJ: They're not doing things completely declaratively
... E.g. a slider which looks like a text box today, a slider if you have an 'HTML5' browser
... Maybe they could add some javascript to get an intermediate behaviour
... I certainly think the XHTML WG think moving stuff into declarative form is part of what they're doing in XHTML2

DC: Not sure what we do next

DJ: Nokia did ask for State of the Art document

DC: Yes, TAG should review the final proposed charter

VQ: Agreed

VQ: Thanks to Dino for his coming to talk to us

Minutes of Last Meeting

<DanC> 28 Jun minutes

DC: says "Draft" all over it

VQ: Title is wrong

<DanC> "Scribe.perl diagnostic output"

<scribe> ACTION: HST to fix all that [recorded in http://www.w3.org/2005/07/05-tagmem-minutes.html#action03]

RESOLUTION: Approved after fixes


DC: Took an action to write a report on sorry state of things

<DanC> OpenID: an actually distributed identity system

DC: Turns out things aren't so bad
... OpenID Looks promising, not yet sure it's a complete solution
... Pain point is blog spam, anyone can add irrelevant or unpleasant 'comments' on a blog
... Idea is you add a pointer to e.g. your home page which in turn provides a secure guarantee that you do in fact control that page

HST: Doesn't stop people doing blog spam, just makes them reveal who they really are when they do

DC: Yes, increases the cost of bad behaviour

VQ: What's plan wrt the action you took

DC: I hope to write a report saying "OpenID saves the day", but not before end of summer, I expect

VQ: End of agenda, AOB?
... Meeting over, next meeting next week same time and place

<DanC> let's see... we have regrets from timbl (among others) next week, right?

<timbl> I gave my regrets at the face-face on the board and in email to Vincent

Summary of Action Items

[NEW] ACTION: Chair to check status of XMLProfiles-29 with Paul Grosso [recorded in http://www.w3.org/2005/07/05-tagmem-minutes.html#action02]
[NEW] ACTION: HST to fix all that [recorded in http://www.w3.org/2005/07/05-tagmem-minutes.html#action03]
[End of minutes]

Minutes formatted by David Booth's scribe.perl version 1.126 (CVS log)
$Date: 2005/07/12 17:13:45 $