#foswiki 2012-02-01,Wed

↑back Search ←Prev date Next date→ Show only urls(Click on time to select a line by its url)

WhoWhatWhen
GithubBot[foswiki] foswiki pushed 1 new commit to master: http://git.io/83Y1PA
[foswiki/master] Item11316: perltidy Render.pm - PaulHarvey
[00:03]
***GithubBot has left [00:03]
FoswikiBothttp://foswiki.org/Tasks/Item11316 [ Item11316: =Foswiki::Render= inappropriately (ab)uses empty =<p>= tags ] [00:03]
gac410pharvey we have great faith in your ability to fix render! [00:04]
pharveyI think I can make it work :)
I'd rather be writing a DOM though :P
[00:04]
gac410will you continue with the regex based parser, or will you need to resort to line-by-line parsing? [00:05]
pharveythe current strategy is to use a multi-pass approach, and simply mark regions of the input as certain types of syntax - each syntax must 'claim' their region of input - and then something magical happens aftewards that assembles a DOM :)
I was going to try to assemble an interval tree data structure, which could be transformed into a DOM tree, but I think having lexed the input, might be best to use Marpa to process the ranges as tokens and build a DOM that way
so in short, there will be regex parsing for some stuff, and line-by-line for others
I have a syntax class called "LineHog", as I hope lists and tables could be done in the same (single) pass
[00:16]
gac410There are some strange things in render that could hopefully be resolved. Some of our wikiword matching can be unpredictable. We use an explicit ending delimiter for wikiwords in some places, and it's open-ended in other places. [00:19]
......... (idle for 44mn)
GithubBot[foswiki] foswiki pushed 1 new commit to Release01x01: http://git.io/fRKijQ
[foswiki/Release01x01] Item11490: Update geturl.pl - GeorgeClark
[01:03]
***GithubBot has left [01:03]
FoswikiBothttp://foswiki.org/Tasks/Item11490 [ Item11490: =geturl.pl= script needs POST capability to run statistics ] [01:03]
GithubBot[foswiki] foswiki pushed 1 new commit to master: http://git.io/RhRZYw
[foswiki/master] Item11490: Update geturl.pl - GeorgeClark
[01:04]
***GithubBot has left [01:04]
pharveyw00t, closer to peeing properly [01:05]
gac410gac410 withholds some comments :-D [01:06]
......... (idle for 43mn)
harlanSo I have not thoroughly read the developer bible. If I have a commit, do I just commit it to the trunk or is there a review process? [01:49]
gac410nope - no process other than the 14 day review for development proposals of core components, or plugin author review if plugin marked "coordinate with author" [01:57]
harlansounds familiar - thanks! [01:58]
gac410Need an open task. Commit message must identify the task. And please keep Trunk and Release01x01 branches in sync for bugfixes or minor enhancements that apply to both. [01:58]
harlanso for the bits I;m talking about (cleaning up language for the Enabled Plugins tab) should that also go in to 1.1? [01:59]
gac410most of us see the commit messages pass by in email, here, etc. You'll be sure to hear from someone if there are objections. ie. To keep Babar happy, make sure you perltidy [01:59]
harlan'k [02:00]
gac410yes. For tab headings, I think there are auto-anchors generated so be sure to peek at the docs to see if the anchor is referenced. [02:00]
harlanI have not done much withb ranches and svn.
Is it expected that the 01x01 branch will pull cleanly into trunk?
[02:02]
gac410hm. well with git, you just checkout a branch and cherry-pick the fixes from the other branch. No idea with SVN - it's been too long since I worked with multiple branches. [02:04]
harlanI have an svn repo - should I convert to git?
harlan considers svn to be somewhat better than CVS.
[02:04]
gac410hm. if you know svn and are comfortable, I wouldn't try to be learning git and foswiki at the same time ;-) [02:05]
harlanI have git stuff to learn anyway, and I have only done basic stuff with SVN. [02:05]
gac410git works well with github, or you can do a git-svn checkout of the entire svn history into your own git repo. I've done the latter.
Read through Foswiki:Development/HowToUseGit for a summary. Foswiki:Development/SvnRepository for svn info.
[02:06]
FoswikiBothttp://foswiki.org/Development/HowToUseGit [ HowToUseGit ] http://foswiki.org/Development/SvnRepository [ SvnRepository ] [02:08]
gac410a full git-svn checkout of trunk is big. Takes a long time.
pharvey has done a lot of work with pseudo-install - it will pull a plugin from github automatically if you try to pseudo-install a plugin that is not in your local repo.
[02:08]
......... (idle for 40mn)
pharveyit's not just Babar that likes perltidy. I find it painful when trying sync up Release01x01, trunk, store2, unicode... and there's a bunch of noise because code formatting is different (makes it hard to see where the real differences are).
not to mention, perltidy'd code in my experience means less merge conflicts.
less/fewer
[02:49]
gac410agreed. Babar is just the most vocal :-) [02:51]
............. (idle for 1h4mn)
***kip3f has left [03:55]
pharveynope. proper <p> tags is just doomed. [03:56]
gac410that sucks. Can you get closer at all - or is it an all or nothing situation? [03:57]
pharveyIt's getting uglier and uglier. For example, I need a robust way to avoid <p> tags from occurring outside the <body> tags, if present. And that's doable. But then I also need to avoid things like <p>\n</div></p>
and for *that*, I'm 99% sure we can't rely on regex hacks.
I'm going to make getRenderedVersion pluggable. That might at least allow the beginning of some progress.
[04:01]
gac410I wonder if regex processing is the right solution anyway. It's not great for performance either I suspect. [04:03]
pharveyMy first thoughts are to (ab)use WysiwygPlugin's TML2HTML.
as an experimental alternative Render impl.
Or maybe as SvenDowideit, dress up the TML with <p> tags as best we can, and then chuck it at htmltidy :)
[04:03]
gac410iirc htmltidy was tough to get installed - might not be a great solution unfortunateloy. [04:05]
pharveyyeah, I know.
Anyway, I think a pluggable render is a good start. Especially to try to allow non-HTML renderings.
[04:05]
gac410yeah - that might be the ultimate solution for pdf, xml. etc. [04:06]
pharveyThere's no easy fix. So a really huge, earth-shattering fix, might benefit from being tentatively and iteratively developed in parallel with the legacy renderer, rather than a complete all-or-nothing switch. [04:07]
gac410yeah. Sites that are dependent upon quirks of the old historical render can continue with it - new pluggable render generates correct html, etc. [04:08]
..... (idle for 24mn)
GithubBot[foswiki] foswiki pushed 1 new commit to master: http://git.io/o0CDFQ
[foswiki/master] Item9551: Add some initial unit tests - GeorgeClark
[04:32]
***GithubBot has left [04:32]
FoswikiBothttp://foswiki.org/Tasks/Item9551 [ Item9551: ImmediateNotifyPlugin has numerous issues ] [04:32]
.................... (idle for 1h38mn)
***gac410 has left [06:10]
..... (idle for 21mn)
pharvey has left [06:31]
............... (idle for 1h10mn)
CDotCDot hopes SvenDowideit is being more supportive than pharvey :-( [07:41]
SvenDowideitCDot, no [07:46]
CDotCDot notes that his proposal is fully compatible with Rafael's original, which is why he didn't raise a new proposal. [07:47]
SvenDowideitin fact, I can't understand why every shortsighted tom-dick-harry thinks they're cleverer than the standard
or why it keeps coming up as the 'solution'
when there are trillions of other standard complying options
my general pov is:
[07:47]
CDotX-error is a *non-standard*, whereas 401 is the HTTP standard for dealing with auth failure [07:48]
SvenDowideitif the standard writers have shown themselves too scared to update the spec to allow the use of 401 for anything other than basic or digest auth [07:48]
CDotCDot recommends careful reading of *the standard* [07:48]
SvenDowideitthen i trust there's a really really good reason
we did that
last time
and _YOU_ concluded that 401 was wrong
and bad
go read the task
[07:48]
CDotCDot can't find a reason, and notes that when people do this - and there are many, many cases - there are no standards trolls complaining [07:49]
SvenDowideitall _I_ did was refuse to allow 400 [07:49]
CDot(07:49:08) SvenDowideit: and _YOU_ concluded that 401 was wrong - wrong. I agreed it was wrong for the *interactive* case. Not the ajax case. [07:50]
SvenDowideiti simply don't see a reason to not use X-something in response, if you're relying on XHTTP....in the request [07:50]
CDotfrom all my reading, i can't find anything saying 401 is wrong (or that anything else is righter) [07:50]
SvenDowideitdid you read your previous conclusion from your previous reading
as you had that opinion first
[07:51]
CDotthe reason for using 401 is that you would have to use some *other* error code for the JS to pick up the error [07:51]
SvenDowideitand then re-re-re-vised your opinion upon further reading [07:51]
CDotso what do you propose instead of 401? 500? 403? [07:51]
SvenDowideitno [07:51]
CDotno? [07:51]
SvenDowideiti propose that we all conculded (sadly) that 200 was the only correct response
we all agreed that it sucked, but that it was more 'correct' and thus safer
[07:51]
CDotCDot is working with a jQuery plugin, and can't *see* the ajax call, can only deal with errors [07:52]
SvenDowideit__did you re-read the task__?
you can, as my eg showed, deal with success
and test for error there
[07:52]
CDot(07:52:10) ***CDot is working with a jQuery plugin, and can't *see* the ajax call, can only deal with errors [07:52]
SvenDowideithell, from there you can throw another event, if you care [07:52]
CDotand my case is typical, I fear [07:52]
SvenDowideitand using 401 breaks other tools
___did you reread the task__ FFS
this _does_ need to be a developer faq
[07:53]
CDothttp://foswiki.org/Development/ImprovedRESTSupport has no linked task. Which other task are you referring to? [07:54]
SvenDowideitthats why i keep asking you if you read the task
i'll google for you then
[07:54]
CDothttp://foswiki.org/Tasks/Item10566
is irrelevant
[07:55]
SvenDowideitThis means that the 200 response code was correct, insofar as we do not attempt to implement the whole HTTP. [07:55]
CDothttp://foswiki.org/Tasks/Item9184 is the interactive case [07:55]
SvenDowideitwould be a quote from you
http://foswiki.org/Tasks/Item1029
would be where you had one (more convenient) opinion
and then later thought about the actual wording and horrid consequenses of the spec
imo it applies even to the non-interactive case
[07:55]
CDot1029 relates to bin/login, which is out of scope here AFAICT [07:57]
SvenDowideitNote that in the case of an RPC (REST) call demanding authentication, the headers returned must include WWW-Authenticate with a 401. We have insufficient support in Foswiki::Func for building correct returns. [07:57]
FoswikiBothttp://trunk.foswiki.org/System/PerlDoc?module=Foswiki::Func [ (Foswiki login) PerlDoc ] [07:57]
SvenDowideitwould be a relevant thing
but thats presuming you did as i asked
and read the task
[07:58]
CDotCDot juyst did, and sees nothing that runs counter to 401 for an XMLHttpRequest [07:58]
SvenDowideitSvenDowideit remembers that you were of that opinion, but no being forced to think further you conlcuded that it was wrong [07:59]
CDotas MichaelDaum says there "For REST calls, a HTTP 200 containing a login screen is a problem, and will always be" [07:59]
SvenDowideitmeans you didna docco your thoughts well enough to argue either way :( [07:59]
CDotI know from looking at the code that *none* of us went into it deeply enough to understand *why* a 200 was returned [08:00]
SvenDowideityes, throwing a html login screen to anything that is asking for a response in json or xml is so fucked up and stupid its crazy [08:00]
CDotI realised yesterday that it was *my* laziness when we were working on the login managers
right, so the discussion is "we know 200 + a login screen is wrong, so what else"?
[08:00]
SvenDowideitbut 200 is not the problem, the problem is that we're not actually obeying requested reposnse type
no, ignore 200
[08:01]
CDotnow, you argure for 200 + JS analysing the respponse [08:01]
SvenDowideitthrowing "html_ when you are asking for json is wrong [08:01]
CDotI'm saying "you can't do that, because so much JS hides the ajax event behind layers of JS" [08:01]
SvenDowideita tmpl auth could just as easily reply anything
and i'm saying that ajax is not the only client of rest
[08:01]
CDotagreed; like 401 + an appropriate WWW_Authenticate header [08:01]
SvenDowideitso you can't presume that you have control over what happens when a 401 comes back [08:02]
CDotI know; that's why we checked for X-Requested-With [08:02]
SvenDowideiter
lots of non-js use that
[08:03]
CDotI *know* that doesn't isolate "ajax" (what can?) but it *does* identify a call coming from JS [08:03]
SvenDowideitno, it does not [08:03]
CDotwhat else is identified that way? [08:03]
SvenDowideitthere's a huge pile of C# that i dealt with that says X-Requested-With: XMLHttpRequest
and pretty much anyone else thats getting around people assuming stuff
if you're talking a custom reponse
should you not be making a xustom request header?
otherwise you're presuming lots of things you can't know
[08:03]
CDotI started down that route, but pulled back when this apparent "standard" emerged
CDot is basing "standard" on wikipedia, *not* a valid RFC
[08:05]
SvenDowideitSvenDowideit shudders at 'standard' wrt rest [08:05]
CDotnothing in the RFC's gives any help (that I can find) [08:05]
SvenDowideitcos auth in rest is idiocy
the entire REST thing presumes that you authenticate before you do something that needs auth
[08:05]
CDotyou can't just throw you hands up in horror and run away [08:06]
SvenDowideitwhich has pissed me off before too
i'm not, i'm just more conservative about deciding i know better than the people that are repeatedtly bugged about fixing that bit of the spec
the browser auth stuff hasn't improved in ~15 years
[08:06]
CDotok, I'm going to take a look at what Rails does here, cos that is treated as a "golden example" by most folks [08:07]
SvenDowideitoh god
rails is consided the canonical eg of the worst idea :)
[08:07]
CDotgive me an alternatve that doesn't involve running away crying?
CDot wants to fix this
[08:07]
SvenDowideiti would be tempted to work out if we can impl digest auth
but not enough to do the research to find out 'no'
[08:08]
CDotCDot isn't sure how that helps, given that digest auth presents the same problems in an ajax context [08:09]
SvenDowideitSvenDowideit looks to see if WWW-Authenticate: yeah-like-fuck is actually a reasonable idea [08:11]
***ChanServ sets mode: +o MichaelDaum [08:12]
SvenDowideitas thats what you're proposing
and is really the root of what niggles at me
SvenDowideit starts to wonder more
[08:12]
CDotright. There's a supposition because the standards only talk about basic auth that only basic auth can use it. But that's *not* what the RFCs say (though I *am* reading between the lines here, I know)
and the impl I proposed - sure it can be improved - works in all the browsers I tested (though I didn't test curl/wget yet, so other user agents may not concur)
[08:13]
SvenDowideitits the 'other agent's and caches or weird proxies that interest me [08:15]
CDotproxy auth is done using other headers [08:15]
SvenDowideiti don't mean proxy auth [08:15]
CDotyou mean a proxy might filter the headers? [08:16]
SvenDowideiti mean proxies that filter [08:16]
CDotCDot wondered about that, but could find nothing online [08:16]
SvenDowideitits a secret world of pain
well, i know of NTLM, negotiate and just found there a Compuserve
http://stackoverflow.com/questions/940206/what-should-i-pass-for-the-www-authenticate-header-on-401s-if-im-only-using-ope
certainly seems desirable
what happens if the browser gets that 401 response ?
SvenDowideit would be happy if this was a correct answer - even more so if we could use status: 401; WWW-Authenticate: html-form for tmpl-auth
[08:16]
CDotyou mean what happens if a browser marks a plain reqest with X-Requested-With=XMLHttpRequest?
cos that's the only way a 401 gets sent
[08:20]
SvenDowideitbut by golly we'd be more comfortable if we could find a little more supporting evidence
no, i mean if you take out that if(X-Requested...
[08:20]
CDoterm, I found no evidence that 401+HTML form is kosher [08:21]
SvenDowideitdoes the browser get all pissy?
there's sod all evidence for any of this
[08:21]
CDotthat was what we were doing before, so I would say "yes" [08:21]
SvenDowideitlots of talk, no actual w3c fixing it [08:21]
CDotno, not that I can find.
but all frameworks have to deal with this, so should be examples aplenty
[08:21]
SvenDowideitso we have no evidence that 401+html is _not_ kosher [08:22]
CDotand BTW, AFAICT C# emulates XMLHttpRequest and uses the header there
no; I did implement 401 + HTML and IIRC it *did* work
[08:22]
SvenDowideitso long as we make WWW-Auth: something_custom [08:22]
CDotbut I didn't revisit it yesterday [08:22]
SvenDowideiti thought that before eugene's code to 400 we had 401
but then we went to 200 due to the discussion and your analysis
[08:23]
CDotyes; I implemented 401 originally, but got flamed and backed down. [08:23]
SvenDowideit(i only picked a random number to kick the discussion off, as 400 was a POS) [08:23]
CDotCDot deferred to people who seemed to know better, and still wil. [08:23]
SvenDowideitno, you pretty solidly supported 200 in the task [08:24]
CDotfor the non-XMLHttpRequest, i think I still would support it [08:24]
SvenDowideitand as all we can (historically speaking) can go on, that task is impotant [08:24]
CDotas expecting a browser to do HTML on the back of a 401 seems wrong to me [08:25]
SvenDowideitreally?
(i'm going to posit an idea)
status:401
www-auth: html
..... thus defines that the payload is html and should be rendered
[08:25]
CDotwell, it makes sense, but there's no browser support for that AFAUK [08:27]
SvenDowideit(seeing as we have some strong anecdotal evidence that html with 401 has er worked?) [08:27]
CDotI think it worked on some browsers, but not all
CDot is dredging his memory
CDot tries it
[08:27]
SvenDowideitSvenDowideit rewinds to : wasn't it working for many many years in tmwiki
until we changed it
[08:27]
CDotnot sure. [08:28]
SvenDowideit(i'm truely not sure, just thought we'd just agreed that is what the code was doing until eugen changed it) [08:28]
CDotit appears the discussion was on the mailing list, not in Tasks
CDot is reading http://foswiki.org/Tasks/Item1029
[08:29]
SvenDowideitoh ffs. i re-found this again : http://www.peej.co.uk/articles/http-auth-with-html-forms.html
not really relevant atm tho
[08:30]
CDotCDot doesn't see that as relevant (or even correct)
CDot can't find any discssuon on the foswiki-discuss mailing list. Wrong mailing list, perhaps?
[08:32]
SvenDowideit:/ f-svn
really need to add the archives as a foswiki web
and then add *search
i wonder tho
[08:35]
CDoty, f-svn, 12/02/10, thread started by gac410 [08:37]
SvenDowideitSvenDowideit much prefers the idea of one status and response to rule them all
but is worried that you, i and other ajaxiers love the simplicity of doing what we want and are blinded to the risks
SvenDowideit reading http://webmasters.stackexchange.com/questions/24443/should-i-return-a-http-401-status-code-on-an-html-based-login-form
[08:39]
CDotat the end of the day, navigating around the limitations of the browsers is going to define what can and what can't be done
CDot notes that Lynx and IE both have reported issues with WWW-Authenticate headers
M$ are even using the WWW-Authenticate header *in the wrong direction* i.e. sending an incorrect WW_Authenticate to the server breaks IIS
[08:42]
SvenDowideitso, at, we 302 to login [08:44]
CDotwhich isn't *disallowed* by the RFC, but sure as heck isn't recommended
no, we 200; we use an $exception->generate with a 200 status
[08:44]
SvenDowideitand then 200 the login
SvenDowideit just asked curl to edit my page
[08:44]
CDoterm, I could be wrong; sec [08:44]
SvenDowideitgot a 302
and then 200 login
the problem is that ajax in js follows the redirect transparently
[08:45]
Babaryeah... [08:45]
SvenDowideitsame as wget does - which is one big bad thing about wget :) [08:45]
Babarwe alreaady had this discussion 4 years ago, didn't we? :) [08:45]
SvenDowideiti like curl -I :)
Babar, thats the problem
[08:45]
CDoty, 302, $session->redirect( $url, 1 ); # with passthrough
in TemplateLogin.pm
[08:46]
SvenDowideitSvenDowideit has to go :(
back tomorrow probly
laters
[08:46]
CDotciao
Babar: we had a very similar conversation, but not exactly the same
[08:46]
BabarSvenDowideit: wget -s, but it's not exactly the same. At least you see the 302 :) [08:46]
CDotand we discussed it aain, 02/10, and I wasn't as engaged as I should have been :-( [08:47]
BabarCDot: the one which ended up in Eugen "fixing" the 302 in being a 401
and then Sven thought it was rubbish, and put back a 200
or maybe I'm confusing this with a very similar issue
[08:47]
CDotsort of. I'm not trying to change that 200, that's a side discussion
I'm proposing re-instating 401 for ajax requests
cos 200 is *useless*
[08:47]
BabarI thought we had that already [08:48]
CDotno [08:48]
Babaroh
Sven: bad boy!
Sven said back then that we should do that
[08:48]
CDotnot his fault
it was an all-round failure to understand and follow up
[08:48]
Babarbut ajax scripts call rest, right? [08:48]
CDotruight [08:48]
Babarso rest can return a 401 login instead of a 200 login, can't it? [08:49]
CDotit can now.... [08:49]
Babarawesome! [08:49]
CDothttp://foswiki.org/Development/ImprovedRESTSupport [08:49]
***Babar sets mode: +oooo AndreU CDot MartinCleaver terceiro [08:49]
CDotbut I'm defending a bit of a flame ATM
CDot goes for breakfast; back in a bit
[08:49]
BabarBabar goes to meeting [08:50]
CDotCDot concludes that he has always supported the "401 for rest" approach, albeit with a brief diversion into 403 which he knows was wrong [08:51]
.... (idle for 15mn)
SvenDowideitBabar, owrse
eugen made it 400
which is the only thing i strongly objected to
i don't like different reponse for rest / ajax, but ... there's a wiggle
[09:06]
.... (idle for 15mn)
harlanHas anybody looked in to why Main.WikiUsers sometimes "loses sort order"? [09:22]
Babaryeah... I saw that yesterday with the Brazilian wiki... the WikiUsers topic was a mess [09:24]
harlanalso fw.o
We've been seeing this for Years though, on a to site.
[09:24]
Babardid anybody open a bug report? :) [09:25]
harlandon't recall [09:25]
...... (idle for 26mn)
I'd like to point USERSWEB at a new web (ie, not MAIN). It's no fun searching for USERSWEB on the wiki. Offhand,, might somebody know where I should look? [09:51]
CDotI did some more reading; there's nothing more authoritative than stackoverflow when it comes to handling 401 in AJAX :-( [09:53]
.... (idle for 17mn)
***ChanServ sets mode: +o OliverKrueger [10:10]
..... (idle for 23mn)
gmcis this a typo in the apache config generator, or do i really need that \ at the end?
PerlRequire "/var/www/wiki/tools/mod_perl_startup.p\l"
[10:33]
CDotgmc: my money would be on a typo [10:45]
gmcindeed so, as with the \ apache refuses to start up [10:45]
fixed it on the wiki.. but, is this a change that needs to go into svn ? [10:58]
..... (idle for 21mn)
harlanWhere is the home/default web set?
I figure it will be safer/better to make a new home/default web that it will be to change the UsersWebName...
[11:19]
CDotharlan: in =configure= [11:25]
harlanwhich section?
I've been thru all the tabs and don't see it, unless the default web is the UsersWebName...
[11:26]
..... (idle for 21mn)
***ChanServ sets mode: +o MichaelDaum_ [11:51]
..... (idle for 24mn)
CDotharlan: {UsersWebName} EXPERT setting in "Login" [12:15]
harlanCDot: I'm seeing that one should avoid changing that as it may affect upgrades. I figured it would be easier if there was a way to change the default web. [12:27]
CDotdefined "default web"? Do you mean the landing web? [12:27]
harlanwhat one sees if one just visits the top-level URL.
So probably the landing web.
[12:28]
CDotright. Well, I made a proposal and wrote some code to make that simpler not long ago. [12:29]
harlancool, so I should just wait? [12:30]
CDotno; IIRC someone bounced me to a plugin that does the necessary
CDot can't remember what it was called
harlan: http://foswiki.org/Development/MakeStartPageCOnfigurable
[12:30]
harlanthx... [12:33]
CDothttp://foswiki.org/Extensions/HomePagePlugin [12:33]
harlanharlan reads that next... [12:34]
I'll give that a shot - thanks! [12:39]
Anybody know wny Item10868 was closed? [12:51]
FoswikiBothttp://foswiki.org/Tasks/Item10868 [ Item10868: HomePagePlugin has an invalid Config.spec value type ] [12:51]
harlanI see no comment there... [12:51]
Babarharlan: http://trac.foswiki.org/changeset/11889 seems to be clear enough [12:51]
harlanharlan didn't see that... looking now. [12:52]
Babarvery bottom of the form, it has a commit
it was just a typo, so nothing serious
[12:52]
harlanOK, then I must be seeing a new issue.
The SiteDefaultTopic is originally empty, and under the box, in red, I see: .spec ERROR! TYPE **T** could not be 'use'd
If that's a bug I can open a task on it after I wake up (it's nearly 0500 here)
[12:52]
Babarit certainly looks like a bug
but maybe check that the changes in the link I pasted are in your version
maybe Sven forgot to release it :)
[12:58]
harlan'k
thx...
[12:58]
......................... (idle for 2h2mn)
***ChanServ sets mode: +o OliverKrueger [15:00]
.... (idle for 17mn)
GithubBot[foswiki] foswiki pushed 1 new commit to master: http://git.io/AyUxJw
[foswiki/master] Item11483: Create local branch if it does not exist - OlivierRaginel
[15:17]
***GithubBot has left [15:17]
FoswikiBothttp://foswiki.org/Tasks/Item11483 [ Item11483: Module repos aren't pushing Release01x01 branch to github ] [15:18]
foswiki_irc9Hi everybody. I have a strange issue that I am seeing with our Foswiki implementation that we are rolling out, dealing with one particular web. When originally created, it was made as a subweb of Main. The web was eventually moved to be a top-level web. As with all webs on this server, we set DENYWEBVIEW=Main.WikiGuest to force logging in (corporate policy). The issue is that attempting to log into this one web at its no [15:27]
gac410What's the issue? Your message was truncated here. [15:28]
foswiki_irc9(let's say the web is called Test, we would navigate to http://foswiki/bin/view/Test/WebHome) returns an error. If I'm logged in, the request works perfectly. The strange part of it is that if I go to http://foswiki/bin/view/Main/Test/WebHome (which no longer exists), I get the login screen. [15:28]
gac410What is the error? [15:28]
foswiki_irc9If I log in at this screen, it succeeds, then tells me that the page doesn't exist. [15:29]
gac410What is the error if you "navigate to http://foswiki/bin/view/Test/WebHome) returns an error" [15:30]
foswiki_irc9The error is something about too many redirects (at least, that is what Firefox says - IE8 just reports that there is an error. [15:30]
gac410Sounds like an apache configuration issue - unless you have an optional plugin installed that does redirecting. [15:31]
foswiki_irc9I've basically narrowed it down to the fact that the login page, for some reason, still thinks the web exists at its original location, while everything else knows it is not there.
Our Apache config is standard, created using the ApacheConfigurationTool (or whatever it is called).
[15:31]
CDotfoswiki_irc9: the login page doesn't "think". It is redirected to if auth is required to satisfy the request *before* any attempt is made to determine if the target page exists or not.
so if the target page does not exist *but auth would be required to view it if it did" then the login page would be seen
[15:32]
gac410What type of login are you using? Apache login, or Foswiki's Template Login? [15:33]
foswiki_irc9Foswiki template login, with LdapContrib being used to authenticate. [15:33]
***alice|wl has left [15:35]
gac410Do you have firebug installed - or wireshark? When you are not logged in and browse to ...bin/view/Test/WebHome for the first time, what is the first redirect that occurs? [15:36]
foswiki_irc9No firebug or wireshark installed (we probably have wireshark on a system somewhere, but it would probably require an act of God to get permission to use it. [15:39]
gac410foswiki doesn't "remember" where any webs are. A web is simply a directory on disk under data directory, containing a WebPreferences.txt file.
You are generally free to move them around - the only issue with moving webs is correcting other links that would refer to the web
Does that web have any special template or skin overrides?
[15:42]
foswiki_irc9Nothing special. I double-checked everything in the directory itself, especially the WebPreferences topic and WebHome topic. Nothing looks out of the ordinary. [15:44]
gac410Can you install firebug extension on firefox? That would help diagnose what the redirect loop is. Your request to get Test/WebHome should 302 OK redirect to bin/login/Test/WebHome?origin=......
The login should not redirect.
[15:52]
.... (idle for 19mn)
***foswiki_irc9 has left [16:11]
profElmHey there, all. [16:13]
ArthurClemensInteresting move: Wikipedia Chooses Lua As Its New Template Language
http://en.wikipedia.org/wiki/Wikipedia:Wikipedia_Signpost/2012-01-30/Technology_report
[16:16]
profElmHey guys, I was wondering what's the best way to modify the add a topic page.
I'd like to have the user create a new type of topic page.
Say...
Create "new client" page, in which there are certain fields that need to be filled in.
[16:20]
ArthurClemensYou would need to create a html form to pass these field values [16:21]
profElmFrom reading, I figured that some sort-of form would be necessary.
I figured that it might be something like a dynamic page template.
ArthurClemens: Are you sure about that?
[16:23]
ArthurClemensif the filled in fields are simple, like user name and date, you can use an edit template instead
so it depends on the data you want to store
[16:27]
profElmIt's just text.
I'm working on a private wiki.
Used in the intranet.
As a result, I wanted my co-workers to be able to go to a web and create a page there.
When they want to create a page (I think FosWiki uses the term "topic"), I'd like them to be prompted to specify certain fields.
Almost like the bug tracker.
[16:28]
ArthurClemensyes, so you need a html form for that
you can write pure html or you can use FormPlugin to help with that
[16:32]
profElmSo, I create that in HTML and then how do I attach that to a "create a topic" option to a web? [16:33]
ArthurClemensgenerally you put that form in a findable place, like a dashboard or WebHome [16:35]
profElmCould I not just change the topic template? [16:35]
ArthurClemensthat would be overwritten with a FW update
you would implement a new WebCreateNewTopicTemplate.txt in a different web than System
[16:36]
profElmAh, I see.
I was wondering if I could go ahead and modify that within the actual UI.
[16:39]
ArthurClemenssorry, that should be the topic WebCreateNewTopic
hmm, doesn't wor
work
[16:40]
gac410ArthurClemens: WebCreateNewTopic is empty - other than set VIEW_TEMPLATE = WebCreateNewTopic So I think all you have to do is create a custom WebCreateNewTopicTemplate in the desired web, which should override the System version. [16:46]
ArthurClemensI tried to sidestep the template stuff. [16:47]
gac410:-) yeah. good point. [16:47]
profElmgac410: I see. Is there any good documentation showing how to create a custom topic?
I'd like to automatically attach a form that I created to the create a topic page.
[16:47]
gac410profElm: Arthur has a good point. Templates are complex. What we do in the Tasks web is use a CreateNewTask topic which prompts for some required information and supplies it to an edit request.
Exactly. Take a look at http://foswiki.org/Tasks/CreateNewTask?raw=on for an example. It provides the defaults for "templatetopic" and "formtemplate" which provide the template and form defaults for the new task.
[16:50]
ArthurClemensdocumentation is here: http://foswiki.org/System/DataForms#Adding_a_form_to_a_topic [16:51]
***ModAcOst has left "Konversation terminated!" [16:57]
profElmHmmm.
So, from the sounds of I, I need to use the <form name='foobar'... method, but my WebTopicEditTemplate is empty.
[16:58]
MichaelDaum_ArthurClemens, lua ... that sounds like a desperate choice [17:00]
ArthurClemenswhat is bad about Lua? [17:00]
gac410profElm There are two very different solutions. template language based Templates, (as done in System.WebCreateNewTopicTemplate) or forms and topic edit templates, like we do with the task system [17:00]
MichaelDaum_been luaing for lighttpd without finding it to be much fun. nor do I see it fit into a templating paradigm, comparing it to jsp or php. [17:00]
ArthurClemensthis is a possible template approach: http://www.savarese.com/software/ltp/ [17:01]
profElmgac410: If I were to go with the topoic edit templates, would the user always have to attach the form? [17:01]
gac410No. Look at the example I gave you.
unfortunately we "overuse" the word Template ... it is terribly confusing. Again, Take a look at http://foswiki.org/Tasks/CreateNewTask?raw=on the hidden <input for templatetopic provides the new task layout. The hidden <input for formtemplate is the default form info.
[17:02]
MichaelDaum_ArthurClemens, ah didnt know that. [17:04]
gac410Arthur - we really need a way to differentiate template types. I still get confused at times. [17:06]
MichaelDaum_lua really feels like simon's basic [17:06]
CDotdoes {Languages}{bg}{Enabled} matter if {UserInterfaceInternationalisation} is false?
gac410: MichaelDaum_: any idea?
[17:15]
gac410My first thought was no, it doesnt. but I really don't know for sure. [17:16]
CDotme neither; that's why I asked [17:17]
MichaelDaum_CDot, not that I know. this array of languages only serves to configure the set of available lexicons.
for instance, it isnt related to the site language or something like that
[17:17]
CDotso, if I set them to display only if user interface internationalisation is enabled, no-one will weep?
(in -configure=)
[17:18]
MichaelDaum_makes lotsa sense [17:18]
CDotk, thanks [17:18]
gac410CDot - we need to be careful about putting too much intelligence into the js front-end of configure. rpm, deb and other package installers really need a cli based configure :-( Though with the current internals that's probably close to impossible. [17:21]
CDotyeah, I know. I'm purely working on the display aspects, buy occluding options that are not "active" when other options are selected
does not affect either what is presented in the UI, not what is saved
CDot will show gac410 in a few minutes
[17:22]
gac410okay - just a thought - otherwise it does sound good. Though one thought - by hiding information, are admins going to miss out on features. (I hated some of the ms word family products that learned seldom used things and hid them for you)
ie - if I never see that my native language is an option, why bother enabling i18n
[17:24]
CDotno, because the dependencies are encoded into Foswiki.spec e.g.
# **STRING DISPLAY_IF ! {UseLocale} || ! {Site}{LocaleRegexes} **
# If a suitable working locale is not available (i.e. {UseLocale}
# is disabled), OR you are using Perl 5.005 (with or without working
# locales), OR {Site}{LocaleRegexes} is disabled, you can use WikiWords with
# accented national characters by putting any '8-bit' accented
# national characters within these strings - i.e. {UpperNational}
# should contain upper case non-ASCII letters. This is termed
# 'non-locale regexes' mode.
# If 'non-locale regexes' is in effect, WikiWord linking will work,
# but some features such as sorting of WikiWords in search results
# may not. These features depend on {UseLocale}, which can be set
# independently of {Site}{{LocaleRegexes}, so they will work with Perl
# 5.005 as long as {UseLocale} is set and you have working
# locales.
$Foswiki::cfg{UpperNational} = '';
the DISPLAY_IF only controls whether the option is displayed or not, by looking at the value of other relevant options
so "do nothing" options disappear until their controlling option is enabled.
their values are still saved, however.
[17:26]
gac410Okay - so I'm from Kronos - I never enable I18N because who would ever believe that klingon is a supported language. So I'll never see that Klingon is available unless i enable I18n
(no I'm not from kronos and don't speak klingon :-D )
I'm just being contrarian :)
[17:29]
CDotPatch to play with at http://foswiki.org/Development/HideIrrelevantConfigureOptions
well, graying out options - GRAYOUT_UNLESS instead of DISPLAY_IF would be trivial to do now.
[17:32]
gac410yeah - that becomes a judgment call. Would visibility of a particular option trigger the user to want to use the controlling option. more a marketing type decision than technical I guess. Some - like all the rcs commands, probably not.
gac410 grasping for practical example. Maybe... Email userid/password... (hey, if I use Net::SMTP then I can sign-in to my email provider)
similar to deciding which options are "expert" -
[17:35]
CDotright. EXPERT is too clumsy and obscure; this at least has the virtue of relating options together.
CDot only implemented EXPERT because it was the least work he could think of to do at the time
[17:43]
gac410It's always a challenge when adding a config variable. should this one rise to the level of being "expert" and hiding it from normal availability.
Don't want to add noise, but some are valuable. Hide or Grayout sound like reasonable alternatives.
[17:44]
CDotwell, it should be fairly obvious from the patch how to extend it to support other degrees of visibility, if you want to experiment. /me welcomes other ideas! [17:46]
ArthurClemensmost of the times graying out leads to clutter, parts of the interface are there but inaccessible without a clue how to activate it. so hiding is better. but completely hiding means that when searching for a setting it cannot be found.
we can also think of a search interface where you enter the name, it gets autocompleted, you see the setting, and now grayed out because it is normally hidden, with the option to activate the parent setting.
[17:54]
CDotArthurClemens: the settings *can* still be found, as you described, in Chrome
not sure if it's worth doing more.... maybe?
[17:56]
ArthurClemensso how can they be found (without trying out your patch right now)? [17:57]
CDotdunno. The can't in FF, AFAICT [17:57]
gac410Ctrl-U and search the raw html? :-D (not very practical but been done before) [17:58]
ArthurClemensthat chrome interface needs to be programmed in a web page [17:58]
CDoty, i see it's not there by default [17:58]
ArthurClemensthe Chrome Settings page screenshots only show a way how to deal with it [17:59]
CDotI *like* what you showed.... and we could (for example) use tooltips to achieve the same thing
might be better to impl. our own search (no, I have no idea if that's possible)
[17:59]
gac410I figure that as long as we are very selective about what we choose to hide. then not worth a lot of work. Hide it *only* if it would be unlikely to trigger interest in the controlling option.
hm. Discovered a while back that checkers don't even run for HIDDEN options in the .spec files.
[18:00]
CDotthe DISPLAY_IF code has absolutely no interaction with checkers; it's a purely visual thing
in fact, I think if you hide an option with a failed check, you may still see the failure message!
[18:02]
gac410understood. Good that would be very important.
Same holds true for expert options.
they "appear" when there are errors.
[18:03]
CDotCDot isn't 100% sure about that, and can't find an option to test with [18:04]
gac410I'm pretty sure that is true. i've used it before.
gac410 tries to remember an example
[18:05]
Yes - put a : in the AuthRealm option and choose md5 passwords and you get an error and the expert parameter appears.
And that exposes a bug. I deprecated "md5" and changed it to "htdigest-md5" but checker only looks for "md5"
[18:11]
CDotCDot has run out of time; back 2morrow [18:14]
gac410bye [18:14]
profElmHm...
Okay.
I have a few questions.
And was wondering if you guys can go over my code.
[18:17]
I'm having a hard time understanding a few lines in the top of the wiki text for the Bug Notification form. [18:28]
gac410Is your example available to the internet? or please pastebin if not. [18:29]
profElmhttp://pastebin.com/embed_iframe.php?i=Frirg4u9 [18:30]
gac410and your questions? [18:30]
profElmWell, I wanted to run through that code.
For example.
[18:32]
gac4101) <input type="hidden" name="topic" value="Client Name" /> that needs to be a valid topic name for the new topic. [18:32]
profElmOn submit, I just wanted to call up the form that is stored in the Main web, which is titled clientIntel form.
That's another thing. Is there a way to get the input from the text box in Client Name as a POST and use it for the create topic?
[18:32]
gac410Make the field not hidden. [18:34]
profElmSo take the <input type="hidden" name="topic" value="Client Name"/> into something like. [18:36]
gac410<input type="hidden" name="templatetopic" value="WebCreateNewTopic" /> needs to be a "skeleton" topic that will be a model for the new topic
<input type="hidden" name="formtemplate" value="%INCLUDINGWEB%.ItemTemplate" /> needs to be a skeleton or model of the form attached to the new topic. with any desired defaults filled in.
[18:36]
profElmBy skeleton/model, what do you mean? Should I not be referencing those values? [18:38]
gac410When the user gets the Edit screen, the "templatetopic" will be the model for what is pre-filled into the edit text area. WebCreateNewTopic is probably not what you want. It's empty.
And formtemplate needs to be an example of *your* form if you want the form pre-filled with defaults.
%INCLUDINGWEB%.ItemTemplate is probably not what you want.
[18:39]
profElmLet me show you the form that I've written...
http://pastebin.com/embed_js.php?i=T2Ed5Ehq
[18:40]
gac410You could also just attach the form to the templatetopic and it will be used. and ignore the formtemplate
So is your form working, and can be attached to a topic and you get what you want on that topic?
[18:42]
profElmI have no idea how to even use the form.
I don't know where it exists or how I can use it.
[18:43]
gac410have you read the documentation? [18:44]
profElmNot thoroughly enough. [18:44]
gac410See Foswiki:System/DataForms Start small. Get your form working, and make sure it is what you want when you view a topic with the form manually attached. [18:45]
FoswikiBothttp://foswiki.org/System/DataForms [ DataForms ] [18:45]
gac410Once you can manually create clients with attached client forms, then move on to using that in a create new client form.
So Create a ClientForm which contains the form def. Edit WebPreferences and make it active in WEBFORMS Then create a example client, and attach the form.
[18:45]
profElmgac410: You are so resourceful, I really appreciate your help. [18:48]
gac410Once you have all that working, and are happy with it. You can automate creating new Clients with a web form. I'd recommend using the AUTOINC for the topic names, instead of user input.
What if user name was "deNname" can't have a topic starting with lower case. So use the client name in the form and keep the topic names un-intelligent.
[18:48]
profElmgac410: I have tested out my form and it works! Now, what I'd like to do is automate creating new clients with a webform.
By webform, we're talking about an HTML form and not the form that I attached in the topic, correct?
[18:57]
gac410Okay - create an "example client" (I hate the term template). Lets call it "ModelClient" And put in any ---++ headings you want by default. And attach your form you have working.
Then create a *simple* form just a few lines from your first try: http://pastebin.com/embed_iframe.php?i=Frirg4u9
[18:59]
profElmgac410: I will be back in a bit. [19:00]
gac410You need <form method="post" action="%SCRIPTURL{edit}%/%MAINWEB%/" name="clientIntelForm" id="clientIntelForm">
Make the topic field an AUTOINC, templatetopic is your "ModelClient" and then the submit <input  type="submit" value="Add Client" class="foswikiSubmit" />
[19:01]
....... (idle for 33mn)
GithubBot[foswiki] foswiki pushed 1 new commit to Release01x01: http://git.io/tEiJgQ
[foswiki/Release01x01] Item11495: Checking if : in AuthRealm incomplete - GeorgeClark
[19:35]
***GithubBot has left [19:35]
FoswikiBothttp://foswiki.org/Tasks/Item11495 [ Item11495: ={AuthRealm}= checker doesn't detect htdigest-md5 encoding ] [19:35]
GithubBot[foswiki] foswiki pushed 1 new commit to master: http://git.io/ImwOkA
[foswiki/master] Item11495: Checking if : in AuthRealm incomplete - GeorgeClark
[19:35]
***GithubBot has left [19:35]
........ (idle for 39mn)
profElmI have returned. [20:14]
Okay. I followed this basic tutorial, but have questions as to where to implement step 2.
http://foswiki.org/System/FAQAnApplicationWithWikiForm
[20:26]
.... (idle for 16mn)
Rather, I think that I've got the hang of it. However, the fields in the form are not being populated. [20:42]
.... (idle for 16mn)
Okay, now I just face one issue with post data. [20:58]
.......... (idle for 48mn)
GithubBot[foswiki] foswiki pushed 1 new commit to master: http://git.io/xM904g
[foswiki/master] Item10247: stronger appearance - ArthurClemens
[21:46]
***GithubBot has left [21:46]
FoswikiBothttp://foswiki.org/Tasks/Item10247 [ Item10247: Implementing Base Skin ] [21:46]
..... (idle for 23mn)
profElmStill having a bit of an issue creating a new topic page with a custom form...
In particular, I have an issue where if my entry has a non-wiki name, it redirects me to a whole nother page. However, I have already specified the "onlywikiname" value="off" setting.
[22:09]
..... (idle for 20mn)
ArthurClemensprofElm: you should be able to create non-wikiword topics
does it redirect at topic creation?
[22:30]
...... (idle for 28mn)
pharveypharvey threw more fuel into the Foswiki:Development/ImprovedRESTSupport fire [22:58]
FoswikiBothttp://foswiki.org/Development/ImprovedRESTSupport [ ImprovedRESTSupport ] [22:58]
....... (idle for 30mn)
gac410hm. pharvey - I think I understood where cdot is coming from. The only requirement is that the server generate a correctly formed WWW-Authenticate header. If javascript is "in control" then it is responsible to respond to the header.
The "200" with a login template is the correct response to a end user responsible for filling in a login request. A 401 w/ WWW-Authenticate aimed at a javascript process seems as though it would be equally correct.
[23:28]
pharveyMy understanding is that WWW-Authenticate is meant exclusively for HTTP layer. Which means XHR shouldn't be a special case.
To me it seems we're trying to force a round peg into a square hole.
[23:30]
gac410I need to read it all again. the 401 doesn't specify the mechanism, only that a 401 must include a header in the response that requests a mechanism. In the Javascript case, as long as it implements the requested mechanism, I think it's okay.
our prior implementations we returned a 401, but ignored the whole mechanism part. and that was clearly incorrect.
[23:33]
SvenDowideiti'd much rather that we return the same status for all req's that result in require auth
but :/
and I like the idea of adding WWW-Auth: htmlform
then again, every time i look at this stuff i'm tempted to go hide in a hole, or to nuke w3c
[23:35]
gac410But they are very different types of auth. hm. interesting. htmlform is indeed a mechanism I guess. ;-) [23:36]
SvenDowideitwhat the mech is, is not at all speced, and there's nothing to insist that the browser implement it [23:37]
gac410agreed. [23:37]
SvenDowideitjust that server needs to tell the browser what it wants
the only kicker is some testing
SvenDowideit goes to his imense and canonical collection of all web clients and runs a test
(magic 8-ball says) maybe
[23:37]
gac410I'm trying to figure out the SafeWikiPlugin. It's doing some really nasty things. [23:39]
SvenDowideityeah :) cool innit [23:39]
gac410It inserts </html> when it replaces the first conditional IF.
Which if assert is enabled, generates a junk after </html> fail.
[23:39]
SvenDowideitwow, that sounds odd
SvenDowideit giggles at the creation of experimental toys we don't want
[23:39]
gac410If you turn off asserts - the client seems to ignore the embedded </html> [23:40]
SvenDowideitSvenDowideit goes to corral kids [23:40]
gac410Actually I don't know where we would send a WWW-Auth: htmlform ... since template auth done with a 302 redirect to a login form. Ah. The login form would be returned with 401 requesting htmlform auth. Hm... [23:48]

↑back Search ←Prev date Next date→ Show only urls(Click on time to select a line by its url)