Nick Kew, Sean B. Palmer, Jim Ley, Al Gilman, Wendy Chisholm, Eric Prud'hommeaux, Terje Bless
Charles McCathieNevile
Monday we indentified the following issues:
State needs to be saved so that we can see a thread of actions taken. Many people might make assertions about some bit of content. As the author takes them into account, the content will change. However, we need to keep track of the state of the content so that as it changes, the author is not receiving all of the messages if the messages are no longer relevant (messages either from the tool to prevent it from checking content that has been checked and o.k.'ed by the author or messages from other people annotating the content, i.e. raising issues).
We continued discussion of hash. Decided that ETag wasn't useful.The idea of hash doesn't seem to have gone anywhere. Instead we talked about threads of assertions and referring between testsubjects so that subsequent test subjects would update earlier ones.
We concluded that we need EARL threading although it wasn't clear if it was something that needed to be added to the language or something that would be taken care of by tools. We decided to think about this and discuss more at Monday's meeting.
NK has pulled the plug on some of the tools. No one complained, so NK won't worry about not making them available anymore. Although, still ought to check with CMN since he might be using.
WC took an action a while ago to find out about automatically logging and pubicly publishing those logs through W3C tools. Sent request to sysreq during the chat.
Discussion of what was needed and why. WC and Charles Munat will discuss the UI. Jim and Nick might do some hacking to help. WC will come back with more info. AG interested in the tool and perhaps will also help with UI.
Monday (Jim's bday :) at the regularly scheduled time.
Here is the IRC log. I moved some bits around, separated out some side conversations, etc to make it easier to read.
WC how long do you all want to chat for?
JL Er... pass
NK Until we have nothing more to say - or get too tired?
JL depends how boring I become.
Al (~asgilman@cp158611-a.alngtn1.va.home.com) has joined #er
WC lol
WC hello all
WC and al. :)
WC one too many l's.
AG None too many greetings.
WC at any rate, let's see where we're at in 1/2 hour, although i expect we'll go 1 hour. my summary of last mtg and the issues we collected:
NK Item: valet experimental WAI tools?
NK (some of them have come down: anyone mind)?
NK I'll take that as a No. good.
WC puts in on the agenda for later
WC in monday's mtg we ended with discussion of Identifying what has been evaluated. the main issues identified:
NK Ah, yes, we were making a hash of it
NK I see an imminent need for State in EARL? i.e. visitor asserts "this is a problem" webmaster takes some action. webmaster asserts "problem has been addressed"
JL I'm very interested in looking at querying EARL, but I'm not sure if that might be better for an another time.
SBP That seems like a good thing to put on the agenda, IMO
JL when I've had more time to look at it myself, and read up on the subject.
SBP bung it at the bottom of the stack
WC an EARL query example by libby miller is available
WC question on the table: state in EARL?
AG Start with a log tape model. IETF timestamp standard allows arbitrarily fine decimal fraction of seconds. If you are unsure, keep a copy of what you got and evaluated or at least an MD5 to detect if another is dfferent.
NK earl:fails --> earl:passes after earl:action
wendy wonders how annotea deal with this?
JibberJim believes they just ignore it and let them get orphaned...
WC on monday we talked about hash. w/out resolution.
wendy double checks minutes...
NK I can make two earl statements (before:fails, after:passes), but how to express the relationship?
NK we made a hash of it?
JL I think you're right Nick, it's unreasonable to just remove user comments because an author has corrected them, and it needs some feedback mechanism to let users/evaluators know.
SBP Remember also that hash is not necessarily secure. We may well end up just asserting that hashes are foolproof in context, though
WC we also discussed ETag, although the downside was not all servers serve them. Also, what if the content hasn't been published yet?
JL A Hash only covers about 95% of urls in clj FAQ 5% others change despite the content being basically static.
WC other suggestions from monday:
JL ETag was only about 70% and was pretty well duplicated by Last-Modified so I don't think it's well enough supported.
JL (we could define them all of course, and let the tools pick the best in context?)
NK I don't think we have to be dogmatic about this
NK Leave it up to tools how to generate unique IDs
WC except that one of the reasons for creating EARL was to pass info between tools. if each generating own ids, then we can't do that can we?
JL I think we need a way of interchanging "this is what I evaluated, you can use this to check it's the same thing you're evaluating"
NK Of course we can! Email message-IDs are created with a wide range of algorithms
NK Jim, yes
NK URL+timestamp
JL That is insufficient with current URLS, due to adverts, small permanently updated portions, the habit of putting on "todays date" on every page etc.
AG URL + timestamp does not cover server instability (sniffing)
NK Erm - I assert it does
JL And you say I looked at this 2001-12-10, I have no way of knowing if it's what I looked at on the same date, unless the server is telling the truth - they don't.
SBP Yeah... EARL needs to be discrete enough to say that accessibility is maintained across variations; language negotiation etc.
AG Question is 'why do you care'? If the point is to discern the reason for a difference in outcome, you basically want to be able to diff the two inputs anyway.
NK - if we express content negotiation
NK URL + context
NK Context = timestamp + HTTP headers
WC Al - we don't always want to diff the inputs.
WC That's the point.
SBP N.B. Timestamp should/may be in the HTTP headers
NK Not reliable
JL I don't see headers as sufficient.
NK why not?
JL Experience of trying to identify non changing pages.
SBP "Date: Thu, 06 Dec 2001 00:21:15 GMT" - HEAD http://www.w3.org/
AG I agree that for efficiency one doesn't want to always have to diff the inputs. But all the things that are based on "what I asked for" instead of "what I got" are unsafe. You need to know if the inputs to the two evaluations were different. The thorough mode involves canonicalization as in XML Signatoure.
NK Indeed, URL+timestamp IS sufficient MODULO Vary: response header
NK (and modulo broken HTTP implementations)
ericP (~eric@pfunk.w3.org) has joined #er
JibberJim knows there are lots of broken HTTP implementations.
sbp waves to Eric
NK indeed, but can we work around all of them? That way madness lies
EP Hi all. I've been invited by Wendy to address annotea technical issues.
NK hi. We were talking about State
JL We don't try and work around any, we add in the ability to compare what the server gave us rather than what it said t was giving us in its headers.
WC eric - do you track state in annotea?
NK I can make two earl statements (before:fails, after:passes), but how to express the relationship?
ericP considers...
EP state tracking, annotea has the ability to remember triples associated with a document.
AG I really don't understand the question; the before:after pair is defined to express the relationship, no?
EP it also has the ability to flush and re-parse the "document".
NK Al, I just invented them for conciseness of my question
SBP to express the relationship, you simply state that the test subject is not considered to have changed according to the processor; and that *can* be either of a purely processor oriented thing, or it can be something more grounded such as a hash
EP if you wish to compare before and after states, i recommend hanging the state predicates off of two objects, representing before and after.
NK and if before/after iterates over several changes?
EP beforeId --hadWheelChairRamps-> no
EP beforeId --hadPresentation-> presentation1
AG eric, how do you express the common ancestor of the two object-states?
EP presentation1 --tooMuchFlash-> yes.
JibberJim wonders if we need anything special for a before and after, can we not have a second assertion saying the resource passes whatever it failed before, and the tools would then know it had been fixed?
EP and a similar object called afterId, for example.
WC earl might "cascade" similarly to annotations.
WC we generate a bit of earl on 1st pass. it fails.
EP common ancestor, i would assume they would both annotate some resource
NK jim, no, that loses tracking info
WC generate more earl after a while. says it passes.
EP beforeId --annotates-> http://www.w3.org/
EP afterId --annotates-> http://www.w3.org/
JL Tracking info is there because you have a different testsubject (with a later date etc.)
WC a thread of earl. each bit associated w/a hash to identify "state" of resource.
EP beforeId and afterId may be much more useful if they are urls created from a timestamp
WC bit = bit of earl.
EP threading, yes, you could relate these timestamped ids by threads.
SBP ([ a earl:TestSubject ] [ a earl:TestSubject ] [ a earl:TestSubject ]) a earl:Thread .
WC problem is, I have been assuming that the only reason for multiple bits of EARL to be associated w/a resource would be diff authors. However, we are now talking about multiple bits of earl from a variety of people and times. just a hole in my thinking. not sure if others had that as well. had not considered an earl "thread" before. but seems reasonable.
NK wendy, you only fill in the holes when you start building real-life apps
NK I've recently started that
WC :)
JibberJim thinks we never should've started implementing things :-)
SBP True. We recognized this right at the start, but you just can't tell until you implement
NK valet as a user-feedback - author-response - etc tracking system
WC threading. ok. duh. cool. how do we do it? before/after ids? hash?
NK (in fact, not unlike bug tracking)
WC yes - bug tracking. cynthia shelly keeps saying that's gonna be a great app of EARL.
JL EARL is bug tracking...
NK i've just started
JL So we need evolution/threading, do we have any concrete proposals or should we think on it?
NK think on it - definitely
WC eric - any suggestions? annotea has been threading annotations...
NK (whether or not we also have proposals here and now)
WC yes, it will need some thinking.
JibberJim thought that annotea couldn't annotate annotations yet?
SBP I gave you one: earl:Thread rdfs:subClassOf daml:List .
TB evo/threading: How about "Current State" instead? "Right now there are 5 assertions of Foo; 3 have counter-assertions"
JL Sean how do you specify one in the list?
EP i'm working on a threading sentence but i want to make sure i give concrete examples...
TB I don't really see the need for threading assertions together.
SBP one?
SBP Hi Terje!
xover waves
SBP Hmm... didn't someone do an email threading schema?
NK But email has a well-defined mechanism
SBP Ralph, perhaps?
SBP True
TB email threads are a simple three structure, no?
SBP so are resources
TB (Albeit arbitrarily deep)
AG not if they are done right. Should cover all quotes, not form tree.
TB Al: The email thread? email has no quote cross referencing!
wendy slaps xover around a bit with a large trout
WC oops.
NK hmmm .. is earl:References sufficient?
WC sorry.
TB *ouch* Wath'd I do /this/ time?
SBP ah: http://www.w3.org/2001/03/thread
AG If you manually edit them into References, email does. not auto. Just should, for the purposes of CSCW
NK sbp: how did you find that?
TB Last time I checked, email only has a "This is my parent" referencing. News has more complex reference facilities.
SBP Dunno. I find lots of stuff
NK Indeed, as do some web bbs
WC thread:inReplyTo could become something like earl:before. we won't necessarily reply to earl bits, but add onto them.
WC as nick was suggesting earlier...but i don't think we need earl:after.
TB How do you garbage collect that?
WC just a way to create a backwards chain.
EP is "quote cross referencing" the ability to quote and reference multiple previous emails?
TB ericP: Yes, I think... :-)
JibberJim assumes Sean has some intelligent agents out there doing his searching...
wendy assumes sean *is* an intelligent agent. :)
sbp laughs
EP having a thread relationship between the different "snapshots" or
WC terje - garbage collect earl:before?
EP events is not a burden on the server, only on the UI where the composer has to add the link.
TB wendy: An arbitrarily long string of assertions. When does it stop being relevant?
NK xover: either forever (archive) or use expiry (a la news) ?
EP xover, i think this is a common problem, not just in threading.
WC good point. i guess as eric just said, - it's up to the author of the assertion to link or not to link.
NK Or the author of the tool ..
TB A huge source of headache on USENET is the arbitrary nature of expiery times...
EP arbitrarily long string of assertions, the added cost to having it threaded is the assumption that one can follow the chain all the way back.
WC hmmm, this gets into combining assertions in general, doesn't it? the other issues on the table?
NK xover, yes; and after many years we *still* have that
WC e.g. what if one person makes an assertion at time x about everything between <html> and </html>
TB ericP: No, it's that an assertion becomes forever valid and relevant; a bug tracker ditches resolved reports.
WC but another person makes an assertion at time y about only img id="z"
EP arbitrarily long string of assertions, -- sort of like bank statements that refer to the previous statement. so long as the statement is useful on its own, you can dump old statements after n years.
NK years??
TB How big is "n"?
EP so we want the assertions to be relevent unless there is a later snapshot or event in the thread?
JL The threading is only relevant to a particular tool/use of the EARL, it doesn't need to be inherent in all versions so expiry would be tool dependant
WC it will depend on the state of the resource when the assertion was made.
NK earl:[Updates|Supersedes] ?
TB Or an assertion that a previous assertion is resolved or obsoleted (and can be hidden or removed).
NK jim, agreed
JL e.g if I was using it as a bug tracking format I'd want them forever, but if I was using them for deciding whether to visit a URL only the current would be relevant.
AG whether you want assertions that have not been superceded or all assertions since King Tut is in the query, not the EARL.
EP it is possible that the thread relationship says Updates|Supersedes
SBP { ?x a :Statement; :year [ math:lessThan [ math:differenceOf ([ is os:environ of "YEAR" ] "5") ] ] } log:implies { ?x a :DiscardedStatement } .
WC jim, nick - i don't agree only relevant to a particular tool.
JL I'm still not sure EARL itself needs threading, isn't threading reasonably tool dependant so could be an extension?
NK no, earl wants something
WC Al - good point. the query is a key piece.
EP i think the thread concepts are pretty wide, including the bankstatement analogy i just mentioned.
TB Maybe EARL just needs a suffciantly flexible reference mechanism that this /could/ be tool specific?
WC nick, jim - by saying threading is tool dependent, are you getting at the same point Al made? that it's in the query?
JL With the EARL itself being unthreaded (you can still have multiple testSubjects in the EARL, it's just how they relate which is tool specific
NK bankstatement - continuity if you change your bank?
EP but perhaps you'd want subproperties for mail threading and EARL threading that added extra semntics.
NK earl - continuity if you change tools
JL I can't see exchange of histories being something that is too common a usage, but I might be blinkered.
NK wendy, no, I want threading in EARL, but let expiry/archiving be tool-dependent
JL gets AL, but you still need to define the relationship between two subjects somehow the query isn't sufficient.
EP changing banks - you lose the thread, but there isn't really one that crosses statements unless you go down to the threaded transaction level.
AG You exchange histories to narrow the range of time within which an exceptional event happened.
EP thesis: the cost of threading is only in how much you grow to depend on it.
EP plus a bit in UI establishing the thread when posting the later states.
AG I am thinking that 'supercedes' is a derived property, not settable except by computation. Yes it is necessary to determine when the subject of a later assertion has a supercedes-ready relationship to athe subject of an earlier assertion.
WC it's close to 1 a.m. for several of you. does anyone want to take an action to do more research on threading? write a proposal for how to deal with state changes?
WC or something else?
WC i get the sense that we're either all thinking deeply or are stunned?
NK I'm happy @1am
AG I'm numb at 8: p.m. But then, I'm numb at 8: a.m.
JL I'm thinking it might be a good idea to postpone discussion to let us flesh out ideas and to move on to other areas?
sbp will be up for many hours to come
WC :)
WC ok. then let's leave it at that and come back to it next week.
WC thanks for stopping by eric!!
AG verily!
NK yes, the power of IRC
EP no prob.
WC nick - do you want to discuss valet experimental WAI tools?
NK Only if anyone has something to say.
NK but we're missing chaals
WC what is it you want to discuss?
NK I've pulled the plug on some of them
NK Basically, the XSLT module was too buggy, so I've removed it
NK I reimplemented the stable tool chaals asked for in CGI
NK - but if any of the *other* tools are *also* wanted ...
WC did the xslt module produce EARL?
JL Wasn't Chaals looking to use the XML output which you've now got on Page?
JL I know I was...
NK Possibly - but he also asked for a stable version of the EARL
WC nick - not much discussion on tools.
WC i guess everyone is satisfied for the time being?
NK OK, I'll take nobody's missing what's not there
TB wendy: Did you talk to W3T about setting up logging for #er?
WC xover: doh! not yet.
JL wonders where the trout should be now...
NK ??
WC i log when i'm on, but a publicly available log of all the time.
WC jim: lol
SBP Logging to the Web would be rather helpful; in real time * xover still wonders what that troutwhipping was all about...
NK sounds very fishy * SeanP feels a Monty Python reference coming up...
WC i was looking for the spelling of eric's last name for the minutes, but instead of selecting "whois" on his name, I accidentally slapped you.
WC slaps wendy around a bit with a large trout
WC there. everything should be in order now.
NK wendy, wahtever you're on, can I have some?
WC lol
WC spicy v8
AG It's a tall rock face your facing, there, Nick.
NK oh good
SBP .google Monty Python fish slapping dance
xena Monty Python fish slapping dance: http://www.montypython.net/scripts/fishslap.php
much later in the discussion... WC i've sent a request to sysreq about logging and publishing the logs publicly.
JL It's currently easier to work with Nicks XML output than EARL output.
NK Agreed. XML is operational; earl was experimental
JL Until we're combining EARL, all single tools can be created more simply from the XML, XML tools are ubiquitous RDF isn't.
JL So I work with the XML...
SBP Ooh, I might be able to use the list extension property (daml:nil), and the little built-in that I wrote: math:memberCount
SBP Huh... it kinda works if I put a ":Null" in the list, but it depends upon how you want to collate results. I think that results should be first class objects too, and they may even need context information thrown in with them
JL Was there ever a consensus on pointing to a fragment of a page in EARL, was it just multiple testsubjects with XPointer - that seems very verbose...
SBP Jim: Did you come across Len's work on the BNF-to-XML conversion? * JibberJim is trying to think what the EARL would like for the input to http://valet.webthing.com/xslt/jim.html
JL Er no I don't think I did Sean.
SBP "XML representations of arbitrary languages" in http://www.w3.org/WAI/ER/
SBP an EARL-ized SP report, I guess
NK How would Jim parse that?
NK As for jim's wysiwyg validator, I'd *love* to implement it as an alternative view in Page Valet
JL Then Iron out the bugs Nick, and go for it :-)
NK Round tuits !!!!!
NK Time - or lack thereof !!!!!
JL Indeed!
JL Don't we need a new testSubject for each XPOINTER currently - so say if 20 images don't have ALT defined to point to each of the 20 elements to say that, I'd need 20 test subjects? this seems very excessive.
AG Is there a daml:eachOf?
SBP daml:oneOf
NK damn:allOf?
NK [sorry]
SBP you can iterate through lists to probably achieve the :eachOf effect that you're looking for
SBP So, the only two important issues that I am aware of are state, and multiple test subjects
NK I thought the discussion of state useful. Looked like it could lead somewhere
JL I'm not convinced it's necessary within EARL, but have no real objections to it.
JL (multiple test subjects do almost everything as state can be generally inferred just from that.)
SBP I guess some people will find it more useful than others. Anyone can extend what we have; the question is how much we should do, and how much we should test the genericity of the architecture :-)
JL <testsubject rdf:about="1" /> <testsubject rdf:about="2"><after>1</after> </testsubject> is sufficient isn't it?
JL So it's just multiple test subjects ( which is there already) with just the ability of one test subject to claim it comes after another.
JL (or a list of others if more threading is wanted.)
SBP yeah, correct principle
SBP the threading is more or less independent of the actual EARL assertions, IMO
JL I also feel it's only really relevant in tools that don't (or rarely) export their history of assertions.
JL Wendy, your request for interest in creating the Test Cases, I don't really understand what's wanted.
WC jim - Chris and Josh created a whole bunch of test files. we want wcag wg members to go through the test cases on various platforms and collect a mass of data. Then we want to combine the data. it would be cool to save as much of this in EARL as possible. Then combine results in a report like:
SBP *sigh* statistics
AG Why the sigh?
NK nowt wrong with stats - as such
SBP Just difficult for me to grok at the moment
WC well, but also:
WC then, we have a table of:
JL So it's general purpose reports about all of WCAG's work really then :-)
NK So if Jim or I or someone hacked up a database and a form or two ...
WC ...then i would be one of the happiest folks alive!
NK ... in principle at least
WC i actually want to do the xslt of the form input.
WC or at least help.
WC i've been dying to get involved in an imp, but all this other wcag stuff is keeping me busy. :\
JL So you're thinking form -> earl -> DB
NK ug.. YM you want to do the UI?
JL knows he won't be doing any UI...
WC charles munat has agreed to help as well.
AG XForms. Do model first.
NK That's great, neither Jim nor I care much for UI
WC he's got a 3 week break coming up.
WC i think he'll do UI.
JL The input is the XML testfiles?
WC yes.
JL the output is EARL reports containing result of test, and browser etc. evaluated on, are the tests dependant on the browser?
AG what? The input is what the user concluded on evaluating the testfiles?
WC xml test case info gets generated into html for user to look at, as well as fill in part of the form the user will use to comment on the test.
WC test file dtd:
JL (ie this is in theory accessible but because I'm using RubbishBrowser7.4 it's not.
WC test files:
WC does something similar to ATR:
WC but instead of a windows app, it's a web form.
AG I'm still waiting for when we come up with a troller that can answer a "more like this" query where more is off the Web at large and "like this" is in access-related patterns.
JL wonders why the App wasn't written for IE it would've been easier...
JL they're using web techs MSXML!
WC i assume b/c chris works in C/C++
WC al - yes that would be useful
JL is slavishilly devoted to javascript.
WC :)
WC i don't want to use javascript in this app, tho.
AG self-knowledge is the root of all wisdom
NK we all have our faults
SBP tries to get list generators to work, to little avail
WC i want server-side processing.
WC i.e. here's what i see:
WC user wants to evaluate test case 1.
WC hmmm.
WC considers UI
JL knows javascript is not just a client side language...
AG User has 15 minutes free; tells server she is present in test lab.
NK down, jibber. Down!
SBP Great; it can't grok lists-as-subjects at all. Just wonderful
WC you would use javascript to do the server side stuff? i was thinking of php combined w/xslt.
AG Server picks a test that user hasn't done yet, and is needed to complete user-class X test-case distribution requirements
SBP is thinking of using the obvious, but can't get it to work
JL I would yes, but then I'm biassed...
NK Al, that sounds fairly routine
AG But it's what GL needs. Round out breadth of user X test distribution.
JL It's definately UI and test cases (which we've got) rather than the technical programming parts that are time here.
WC sbp - what's "the obivous"
SBP CWM
WC al - if the server were to select tests, it would have to know: which browser, platform, and AT the person using since we'll have to test
WC each case for each config.
WC (ideally)
NK that's easy - we tell it
WC yes, just another part of the process.
WC making sure it gets included in the scenario.
AG users have registered default profile there; most test subjects only do one. Others can flop personal for different parameter profiles
WC Process
WC databases needed:
JL The first part of speccing this is to get the UI person to spec out the UI and the routing, the implementation is hours.
WC the question is...who is the UI person? :) I'm suspecting that will be me and Charles Munat, and possibly Al Gilman...
WC since he seems interested. :)
NK yep, could do that in one session if an ugly hack was allowed
AG Jim: that's entirely a matter of problem-solving style; others would say the first thing is to describe the relation you want to winde up with in the DB
WC Regardless - let me think about this some more and get back to you.
WC unless - there are more thoughts now... :)
WC i haven't any.
NK Al, I think that's too straightforward to make an issue of
AG Now I need the extra-verbal cues. That's too clever for me to decode. I'm not trying to make issues, I'm trying to get GL the infrastructure they need to review facts.
JL Nick - do you know about the RDF triple database for storing EARL?
NK OK, what I mean is, it would take me longer to spec up the database than to implement the whole thing, but the implementation would be an ugly hack
JL (as used by Annotea currently.)
NK jim, sweet f*** all
JL more than me then...
JL it should really be in that. (how do you go from triple to xml-rdf even)
SBP serialize it!
JL Yes...
NK sbp - did you write a proggie for that?
AG But the problem is that we want to retain a cross relation on (user, test, result) triples under conditions where there are further detailed properties of users, tests, and results that are not stable.
SBP If you want to be clever, you store your NS prefixes in a dictionary. IF not, you just hack off the last char and use that as a namespace
SBP I wrote an XML RDF serialization function, yes
SBP er... use the remainder as a namespace, I meant
JL It's all stored in EARL Al, we could have a single EARL for each report, and then it's just the EARL combining we need to do for that, I don't see the problem you're describing,.
AG Wendy, I don't know if I am any use on the UI or not, but don't fail to ask.
AG Jim, that's OK. The devil is in the ontology that lets us roll up the EARL that has different keywords all through it. Just like our question about when does a new test supercede an old test?
JL One problem on the dev work is me and Nick will have very different approaches, and it's definately one of us do it I'd've thought, we'd get in the way more than help....
SBP calls mod_grammar
SBP tip: avoid SHA512 for cryptography... "*** AaronSw changes topic to 'We're building a decentralized SHA512 -> plaintext service. Watch out!'"
NK :-)
AG Can we have a session on support to WCAG WG?
JL Let me know about anything that's discussed later won't you, do we have a date for the next one?
WC i think we ought to talk again next monday at the regularly scheduled time. jim - you'll be off celebrating your bday by then. :)
WC we'll take minutes, tho and you can contribute to the list.
WC if you like, we could schedule time again next wednesday as well?
JL 15:00 no I'll be in the office.
$Date: 2001/12/09 02:06:22 $ Wendy Chisholm