See also: IRC log
Scribe: Ed
agenda: http://www.w3.org/2001/tag/2006/08/08-agenda
Discussion regarding Minutes of 18 July, 2006.
RESOLVED: to approve minutes 18 July
RESOLUTION: Minutes 18 july approved.
<DanC> Note some actions done: VQ sent quarterly summary; Noah sent reply to AB on Use of Unregistered Media Types in W3C Recommendations"
<dorchard> http://www.w3.org/2001/tag/doc/versioning-20060726.html
issue XMLVersioning-41
<DanC> partial understanding bit is new http://www.w3.org/2001/tag/doc/versioning-20060726.html#div437044120
Dave: would like to run through some additions that I've made.
DO: if you take a look at section
1.1.1 compatibility
... I've added a section called partial understanding.
<DanC> hmm... "Partial Understanding a language is the creation of a language" looks wierd. understanding is creation?
<Noah> I had the same problem
DO: defining backwards compatible
discussion on the definiation of backwards compatible.
DO: I think there will be some tweeking of the definitions, but I want to make progress on the notion of capatibility.
<DanC> I think the backward/fwd compatibility definitions need to refer to the mapping from syntax to semantics, which seems to be called "Act of Interpretatin" in http://www.w3.org/2001/tag/doc/ext-vers-generic-uml-v4.png
<Zakim> Noah, you wanted to say that forward and backward compatibility should be symmetric wrt both syntax and meaning
<Noah> My preferred approach, as I've said before, is to start with partial compatibility, and view full forward and backward as the interesting special cases.
<DanC> (hmm... I'm trying to figure out whether to have Dave present and defend each definition, especially since Dave said he doesn't claim the defns are all right.)
<Noah> I think we should be asking: when someone with knowledge of language 2 receives a document sent with knowledge of language 1:
<dorchard> (would prefer to go through the doc, then return. allows me to get to some interesting conclusions)
<Noah> a) Which if any such texts will be rejected as illegal syntax
I think the definitions in the doc may add too much complexity to an otherwise easy topic.
<Noah> b) Which will be accepted and yield information which is clearly different, perhaps dangerously different, from that which the sender intended (e.g. I sent true, you read false)
<Noah> c) Which information will be conveyed exactly as intended (family name is Smith)
<Noah> d) Which information is partially conveyed (this guy has a last name, but I can't grok what)
<Noah> I would build the story >starting< with that.
DO: I think it would be useful to go through the partial understanding.. and then return.
<Noah> Full backward compatibility is then the special case: a) all texts legal, b) never a dangerous misconclusion c) all information conveyed exactly as intended.
<Noah> Partial understanding should be explained as degrees of the above, with the goal being to do what meets the needs of the application, which may vary with circumstance (this application may be OK just knowing reliably that the person has a given name)
<ht> Wrt the current discussion, I highly recommend an ancient document by Eve Maler, written as input to the Schema WG work on open content models, "Substitutability Requirements for Different Applications": http://www.w3.org/XML/Group/1999/08/s14y-paper
<ht> Also "Refinement Task Force Report": http://www.w3.org/XML/Group/1999/07/schema-refine/refinement-taskforce-report.htm#AEN731
<Noah> BTW: Filling in the formulation I started above, Full Forward Compatibility would be exactly the same as backward, but with the documents sent from someone with knowledge of newer language to older.
<Zakim> ht, you wanted to mention this old work
<dorchard> ht, I really liked Eve's paper
<Noah> I'm starting to feel that our process for discussing this isn't quite as effective as it could be. For future telcons, I suggest that we try to review only drafts that have been out for a week or so, and that if a roadmap is needed, that they should be sent in accompanying email. That way, we can just ensure that everyone on the call has read drafts and roadmaps, and start discussion from there.
<DanC> (is it world-readable now? the /Group/ suggests not)
<Zakim> Noah, you wanted to go through my proposed model of partial understanding as logged above
<ht> That paper is member only
<ht> DO, did Eve put it elsewhere?
<dorchard> let me check
<ht> http://www.pacificspirit.com/Authoring/Compatibility/s14y-paper.htm
<ht> You already did it!
<dorchard> exactly!
<DanC> (the question of whether to go general-to-specific or specific-to-general is an art form. I lean toward the way Dave's got it written currently, though I need to read it more closely.)
Noah: I would have said 'given one language, there are many other languages that for some of the instances there is partial understanding.. I'm trying to say that a language is not an understanding.
Noah: to me, this partial
understanding is a more symentric relationship.
... between a sender/receiver and two languages.
... to me it would be very useful to see that layed out in some
kind of firm way and explain in ways that people can
understand.
<Zakim> DanC, you wanted to ask if the good practice notes under "This leads us to two versioning practices" argues in any particular direction about whether css documents should contain
Dan: is this relevant to the discussion of if CSS documents should have version indicators.
<Noah> FWIW: I also think the advice on conservative in what you produce and liberal in what you consume is NOT in all cases good advice. Take XML for example, it does NOT encourage accepting (MYTAG) as a substitute for <MYTAG>, and for good reason
Dave: I don't know if that's
relevant.. because I dont know what they're using the version
indicator for.
... they dont use the version indicator in CSS.
<Noah> The liberal/conservative stuff is a tough tradeoff, between the liklihood that the system will keep doing something, and whether it will in some sense behaive reliably
(Ed thinks CSS should have a version indicator for the very reasons of compatibility)
<Noah> I tend to buy the be conservative in what you send/flexible in what you consume mainly when humans (or some other agent) is there to pick up the pieces when mistakes are made
<Zakim> DanC, you wanted to give a counterpoint to "use least understanding" cuz it leaves no spare bandwidth
<dorchard> Henry, I use the term substitutability because of Eve's paper
TV: The moment you go down the
'people are flexible in what they consume' so I can be flexible
in what I produce than this creates a potential hole.
... We've always had this problem on the web and it appears to
be getting worse.
<Noah> DC: Being liberal in what you accept can dig a hole, because it can make it hard to later give a particular meaning to a TAG that's been "liberally" accepted.
DO: I think the partial understanding and advising people to use the least power is great on the 'consuming side' but we should be conservative in what we produce.
<Zakim> dorchard, you wanted to mention TCP has nothing to do with humans
<dorchard> liberal in accepting works when there's a feedback loop to fix the instances back to the "conservative"
<DanC> (holy grail: explain the interaction of The Internet Robustness principle with "people will take the shortest path to victory" and "follow the money" from micro to macro economics)
<dorchard> (lol!)
<dorchard> TVR: there is a difference between languages designed by humans for humans vs protocols between machines
<DanC> ACTION: DanC to review definitions of partial understanding, backward compatible, and forward compatible [recorded in http://www.w3.org/2006/08/08-tagmem-minutes.html#action01]
<dorchard> Noah: I don't believe there's a bumper sticker, there are trade-offs
<Noah> Exactly.
DO: I'll observe that Noah is pushing back on the bumper-sticker but I've been trying to use the bumper-sticker because I think its useful to think about partial understanding.
Dan: Do we want to take the whole call on this?
DO: this is usefull
<dorchard> TVR, I posted a blog entry on protocols
<dorchard> http://www.pacificspirit.com/blog/2004/06/14/protocol_extensibility_and_versioning
TV: Dave, how do you see the
scope of this document? Is is language only or machine read or
machine consumed.
... it could be 1) purely talking about language versioning
(machine to machine)
<Noah> My concern is that, if you take one side of a tradeoff and make it a bumper sticker, then one has to ask how people learn about the tradeoffs.
2) Purely web (produced by machines, but also consumed by people.
<Noah> If this really isn't a tradeoff, or mostly isn't, then fine: keep the story simple.
DO: its both.
<DanC> (I didn't hear a clear 2nd bumper sticker; I'm inclined to ask Noah to draft it.)
<Noah> If this really is a tradeoff, then oversimplifying it isn't educating our audience to the truth.
<ht> +0 to the tradeoff approach
TV: then I think we should spend time working on the two bumper-sticker idea (Noah's)
DO: yes, I really like that idea.
Dan: so what is the 2nd bumper sticker.
Be conservative in what you accept and conservative in what you produce.
Noah: stick to the speck.
<ht> DanC once said "validate at trust boundaries"
<DanC> yes, I think "stick to the spec; it promotes interop" is a good counterpoint to "be liberal..."
<ht> That's a bumper sticker which amounts to "Be conservative in what you accept"
ht: one of the reasons xml has been so success is because its so strict.
DO: I think this has been really useful
<ht> ht is already on the hook for reviewing. . .
Ed: I can draft the agenda and chair on the 15th then..
<Noah> HT: I think XML is an important example. It's acceptance seems to trace in part to the lack of flexibility in what's accepted.
Dave: I think I'd like to take another shot at it.
<Noah> I have a strong +1 what what Henry said.
<DanC> (nobody is precluded from sending comments/suggestions to www-tag, meanwhile)
<Noah> I should note that I will around for the call next week, then gone for 2 or maybe 3 weeks
<DanC> DO: ETA 3 weeks, i.e. 29 Aug
DO: I can try another revision 3 weeks from now. .. the 29th for publish.
<DanC> DanC: friday before helps
<DanC> DO: ok, I can prolly get it out for review Thu/Fri before
DO: I'll publish on the 24th/25th for review on the 29th.
Noah: I'm at risk for that date.
<DanC> NM: I'm at risk for the 29 Aug call
<DanC> regrets 15 Aug: TBL, VQ
<DanC> PROPOSED: to meet again 15 Aug, Ed to chair, Noah to scribe
<DanC> so RESOLVED.
<DanC> HT: the DI WG has indicated interest in the genericResources draft
<DanC> ACTION: HT to coordinate with Ed on involving DI WG participants on for genericResources-53 discussion 15 Aug [recorded in http://www.w3.org/2006/08/08-tagmem-minutes.html#action02]
Meeting adjourned.
