#foswiki 2012-07-04,Wed

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

WhoWhatWhen
foswiki_irc0I need some help regarding authentication - anyone around ? [01:46]
harlangive it a shot... [01:47]
foswiki_irc0cool.
To start with, I'm tasked with moving a very old Twiki site to foswiki... 2 steps - keep the users and group authentication mechanism
Then after moved to foswiki
the 2nd step (later) we will use LDAP authentication.
[01:47]
pharveyboo! [01:49]
foswiki_irc0My problem is that it *seems* that foswiki requires that I do more than user/group authentication.
Hah!
Yeah, boo!
Powers that be want to keep old authentication until
we're ready for implementing ldap.
[01:49]
harlanthat hsould not be a problem... [01:50]
pharveywhat do you mean by "seems" - are you getting an error message? Or is it just that old passwords arent' working? [01:50]
foswiki_irc0Initial authentication works fine. But when I want to edit a page
I'm asked to login as a user.... It's a different login
Foswiki user login, I guess. (not a lot of experience w/foswiki)
[01:51]
pharveydid you start with a fresh config? Or did you try to make Foswiki work with the old LocalSite.cfg? [01:52]
foswiki_irc0I started by using the tool to generate a config [01:52]
harlanthere is a chance that you have apache auth (for example) to log in at all and that does not pass credentials back to you, so once you have logged in to the site, you have to log in again to make edits. [01:52]
pharveyby "config" i don't mean apache config - although the apache config may be relevant [01:52]
harlanpharvey will know more about this than me... [01:53]
pharveyoh, harlan is right. You might have TemplateLogin configured, but you actually want ApacheLogin [01:53]
foswiki_irc0I used the foswiki config generator to give me a foswiki.conf that I included in my httpd.cxonf
.conf
Oh! Okay, yeah, that could be it.
[01:53]
pharveyLocalSite.cfg is the name of the config file where Foswiki keeps its perl-ish settings [01:53]
harlanor vice-versa... [01:53]
pharveyTemplateLogin -> cookie-based auth [01:54]
foswiki_irc0I think I used TemplateLogin first... [01:54]
pharveyApacheLogin -> lean on apache for auth [01:54]
foswiki_irc0so I'll try ApacheLogin then. [01:54]
gac410pharvey - I answered your question in Item11936. Yes, The intention is to ship the sample .htdigest-configure file with foswiki. [01:54]
FoswikiBothttp://foswiki.org/Tasks/Item11936 [ Item11936: Implement default htdigest protection for configure script ] [01:55]
harlanOne of the things I liked about FW is that for old sites I had to us ApacheLogin(?) first to limit access to the site, then folks could visit but would hvae to log in again to edit. [01:55]
pharveyhi gac410 - I just notice that my checkout of core is always modified now :P [01:55]
harlanUnder FW I only use the normal login, but I make most everything unreadable by WikiGuest, so folks have to log in. [01:55]
foswiki_irc0I like that too, but that wouldn't be backwardly compatible at this point, and just piss off a lot of people (the extra login to edit) [01:56]
kip3ffoswiki_irc0 - I recommend TemplateLogin, because form-based login page is much nicer than apache login (a dialog box from your web browser) [01:56]
gac410Ah. yeah. When you set the password. [01:56]
pharveyon the other hand, if you want LDAP auth, it's almost as easy as just configuring apache rather than Foswiki [01:56]
foswiki_irc0@harlan, thanks. I'll see if that gets me going. [01:57]
gac410git update-index --assume-unchanged core/working/configure/.htdigest-configure [01:57]
pharveylogging in twice is naf :P
gac410: I'll add it to my ignore
(hrm, unless I can't ignore added files)
[01:57]
foswiki_irc0pharvey, when they're ready for ldap, it'll be a piece of cake.
Thanks guys.
[01:57]
gac410Not sure. I've usually used that -assume-unchanged option. [01:58]
pharveyThat's ok. Today I'm writing modbus protocol driver! Yay, fun things :)
pharvey only half sarcastic
[01:58]
SvenDowideitpharvey, modbus - i thought there was an open source impl
iirc even a perl one
[02:00]
pharveythere are several [02:00]
SvenDowideitmind you, if you can use OPC :p [02:01]
pharveyOPC :/ [02:01]
SvenDowideitOPC is fun >:}
FoswikiBot, OPC is fun, it'll save the world!
[02:01]
FoswikiBotSvenDowideit: Okay. [02:02]
SvenDowideitOPC [02:02]
pharveyew
I'm using python-pymodbus
Well, I'm learning, should I say
[02:02]
SvenDowideitgiggle
pharvey, will you be going to lca2013?
i'd love to goto devops-au, but thats next month in katoomba
[02:03]
pharveyI have an MTL-8000 loaded with 8 I/O modules with 96 inputs and exactly one sensor to test with [02:04]
SvenDowideitso i'm struggling to be motivated
hehe
wire up that sensor to all the inputs >:}
tzzzzzt
[02:04]
pharveythis sensor is only compatible with 32 of the channels :P [02:04]
SvenDowideiti've a pile of one wire temp sensors, and an arduino you could spend weeks getting working... [02:05]
pharveyoh wow, lca2013 is in Canberra [02:05]
SvenDowideity [02:05]
pharveyit would be silly for me to not go [02:05]
SvenDowideitproposals close 6th of jul :( [02:05]
pharveyno way I can commit this early :/ [02:05]
SvenDowideitthat too :/
let alone knowing what i'll be doing then
[02:06]
pharveyI'd be happy to hijack a miniconf [02:06]
SvenDowideityup :) [02:08]
harlanSvenDowideit: pm ok? [02:14]
SvenDowideitharlan, always :)
the only questions are how many days before i notice :)
[02:15]
***SvenDowideit sets mode: +oooo AndreU gac410 Lynnwood pharvey [02:16]
gac410Item11981 - I don't understand why USERINFO_cloak is not always set to true or false for each expansion. [02:17]
FoswikiBothttp://foswiki.org/Tasks/Item11981 [ Item11981: USERINFO: broken with FastCGI and {AntiSpam}{HideUserDetails} ] [02:17]
gac410Ah ... never mind. The issue is when the macro is expanded without a parameter. [02:19]
pharveySvenDowideit: thinking of buying one of these http://www.ebay.com/itm/DataHand-Systems-Professional-II-Keyboard-/300734102001?pt=PCA_Mice_Trackballs&hash=item46052639f1
which costs 5x more than my laptop, haha
[02:24]
SvenDowideithehe
pharvey, i liked the idea too
but you need to build oyurself a chair for them
so you don't bend ze wrists the wrong way too
luckily, there is an alternative
think more, type less, produce better systems :p
[02:39]
pharveybut it moves with your wrists?
anyway, more garbage in more faster must mean more garbage out!
[02:40]
SvenDowideitpharvey, tada :p [02:48]
gac410pharvey Item11981, I 2nd your "confirmed" - I'm pretty sure I can recreate in in a unit test too. :( [02:54]
FoswikiBothttp://foswiki.org/Tasks/Item11981 [ Item11981: USERINFO: broken with FastCGI and {AntiSpam}{HideUserDetails} ] [02:54]
gac410Blame me for that code :( [02:55]
pharveyI won't blame anyone for improving code :) [03:07]
GithubBot[foswiki] foswiki pushed 1 new commit to Release01x01: http://git.io/k2PQGw
[foswiki/Release01x01] Item11981: Make sure userinfo cloak is reset. - GeorgeClark
[03:20]
***GithubBot has left [03:20]
GithubBot[foswiki] foswiki pushed 1 new commit to master: http://git.io/GBRFRg
[foswiki/master] Item11981: Make sure userinfo cloak is reset. - GeorgeClark
[03:21]
***GithubBot has left [03:21]
........................ (idle for 1h57mn)
SvenDowideitwho the flying heck decided that i should start sprint training anyway
frickin madness, thats the most exhausting thing i've done all day! all week even
SvenDowideit dribbles off the chair to sleep under the table
[05:18]
................................... (idle for 2h51mn)
***zelazny.freenode.net sets mode: +v FoswikiBot [08:10]
zelazny.freenode.net sets mode: +v FoswikiBot [08:15]
GithubBot[foswiki] foswiki pushed 1 new commit to master: http://git.io/d4fVng
[foswiki/master] Item11055: - MichaelDaum
[08:16]
***GithubBot has left [08:16]
FoswikiBothttp://foswiki.org/Tasks/Item11055 [ Item11055: add support for server aliases ] [08:16]
................. (idle for 1h24mn)
***ChanServ sets mode: +o MichaelDaum [09:40]
.................... (idle for 1h36mn)
CDotinteresting; NYTProf of converting a store from RcsLite to PFS: 99s spent in RcsLite, 9s spent in PFS. Convert PFS to PFS: total time 45s. Convert PFS to RcsLite: Ctrl+C after 10 minutes. This is a command-line conversion.
The RCSLite runtime is mainly in the _readTo method that reads the ,v file
running RcsWrap->PFS now
[11:16]
PeterSvendsenHi all, short question... I have a topic "defectLifecycle" that people can access.... I would like this topic to appear if somebody make a search of the word "serverity"... who can I solve that? [11:20]
CDotcrikey; RcsWrap is even slower than RcsLite (command-line invocation) [11:22]
Colasput "severity" in the body of the topic
and you can hide it from normal view either inside html comments: <!-- severity -->
or via css <span display='none'> severity </span>
(the above for PeterSvendsen )
For advanced use, best would be to add a "keywords" field to the pages, via http://foswiki.org/System/DataForms and put "severity" in it
(fields are stored in page conetnts, so are found by the search
[11:22]
***ChanServ sets mode: +o MichaelDaum [11:35]
PeterSvendsenColas thanks.. regarding the dataform would you have a datafield "Keywords"? [11:47]
ColasI was suggesting it because I am guessing it is your actual use case: you want to "categorize" your pages
I did not suggested the tag plugin http://foswiki.org/Extensions/TagMePlugin as I dont think the standard wiki search uses its tags
[11:59]
........... (idle for 50mn)
PeterSvendsenCola it is only a few pages I would like to do it in.... thanks for your input :) [12:51]
...... (idle for 27mn)
foswiki_irc4Hello Community, we are using foswiki LDAP integration and receive errors when savin a changed topic with LDAD users, whose login name is numerical, eg. 12345, The error is: Foswiki detected an internal error - please check your Foswiki logs and webserver logs for more information. path -m%COMMENT|U% -t-none -w%USERNAME|S% -u%FILENAME|F% of ...path failed: Can anybody help on this issue? Thank you [13:18]
............. (idle for 1h0mn)
MichaelDaumwho has done this ffcache in FormField.pm
the SMELL there at the top is spot on
just got caught by this ffcache not being invalidated on a topic save action
[14:18]
CDotMichaelDaum: it was introduced way, way back, in (tm)wiki. [14:33]
MichaelDaumMichaelDaum fixing it ... http://foswiki.org/Tasks/Item11982 [14:33]
CDotI have never been happy about it, frankly. [14:33]
MichaelDaumit was okay before
but yes, these kind of optimizations belong to at least one level downwards
[14:34]
wow http://jquerymy.com/ [14:40]
CDot"before" is a looong time ago, AFAIK [14:40]
MichaelDaumit was fine before your reorg as the ffCache was properly cleaned up at the end of the request. [14:44]
CDotCDot sees that adding meta-data to webs would be *so* easy with the PFS, if only the store API supported it! [14:45]
MichaelDaumhow about adding PFS to the core? [14:46]
GithubBot[foswiki] foswiki pushed 1 new commit to master: http://git.io/I1pT4A
[foswiki/master] Item11982: fixing broken formfield cache ... this cache never was cleaned up - MichaelDaum
[14:46]
***GithubBot has left [14:46]
FoswikiBothttp://foswiki.org/Tasks/Item11982 [ Item11982: broken formfield cache ] [14:46]
MichaelDaumand show that old lady called RCS the other side of the road [14:46]
jastand add an option for content addressing to PFS :}
several of our wikis have *lots* of duplicated attachments
[14:46]
MichaelDaumCDot, what's the current state of PlainFileStoreContrib? [14:47]
CDotMichaelDaum: working with one caveat; it does not store the author or save comment on attachments
CDot is just reorganising it to do that
[14:48]
MichaelDaumMichaelDaum likes the idea of f* all external version control systems and do it the KISS way [14:49]
CDotI was trying to use the META:FILEATTACHMENT to store meta-data for the attachment, but that just doesn't work (which is why the RCS stores ignore it) [14:49]
MichaelDaumwhy not? it is a meta data record like all the others [14:49]
CDotdid you see the NYTprof results I posted earlier? [14:49]
MichaelDaummissed them. got a pointer? [14:50]
CDotin this channel, at 12:16 GMT [14:50]
gac410gac410 was wondering if for attachments we should save a small _meta.Attachment that has just the rcs-like summary information for each save event. [14:50]
MichaelDaumMichaelDaum would like to add an isthumbnail="true" to image attachments to use this one to visualize search results and user profiles [14:51]
CDotin large batch operations, the PFS is roughly 2X faster than RcsLite (RcsLite running with no suspect topics, that is) [14:51]
MichaelDaum_readTo...yea seen that in nytprof [14:52]
CDotgac410: something like that. I'm saving a "side file" for each topic and each attachment that stores the meta-data for that revision [14:52]
MichaelDaummost of the time it is called reading up to the next semicolon [14:52]
CDotnot sure about the number of files implied by this; but I'm working on the basis of KISS, it can be optimised into a little DB later, once it works [14:53]
gac410CDot: right on ... [14:53]
MichaelDaumwhats the directory layout of PFS? [14:53]
gac410a meta side-file per topic-attachment makes moving of attachments pretty easy. [14:53]
CDotI'm still arguing with myself about that, but the latest iteration is:
* Topics are in data/.../topic.txt. If there is no .txt for a topic, the topic
does not exist, even if there is a history.
* Topic histories are in data/.../topic,pfv
* Each rev of the topic has a numbered file containing the text of that
rev (1 2 3 etc) each with a corresponding metafile 1.m 2.m etc.
* Attachment histories are in data/.../topic,pfv/ATTACHMENTS/attachmentname
* Each rev of an attachment has a numbered file containing the data for
that rev (same as a topic), each with a corresponding metafile (same as a topic)
but I would welcome suggestions before I commit that to code :-)
note that I have deliberately moved attachment histories *out* of pub. This is because in pub they are vulnerable to footpads.
i.e. you can read the history of an attachment by crafting a suitable URL (a known problem with the RCS layout)
[14:54]
gac410hm ... a separate meta file per attachment revision? It was thinking that a single file per attachment listing all the revisions might be faster for things like listing the revisions and comments. ie RCS without the diff [14:57]
CDotwell, I have a test with 10,000 revisions. A single file would be quite slow....
but yes, it's a tradeoff
[14:57]
gac410hm What's the impact of 20,000 files (10,000 meta + 10,000 versions) on the file system? [14:58]
CDot(15:54:08) CDot: I'm still arguing with myself about that [14:58]
gac410ah...
as long as your argument doesn't get violent :D
[14:58]
MichaelDaumhow about movign attachments out of the topic,pfv area [14:59]
CDotMichaelDaum: where to? [14:59]
MichaelDaumdata/.../attachmentname,pfv/ [15:00]
jastit depends on the filesystem [15:00]
MichaelDaumsame as a topic [15:00]
jastsome filesystems have a limit on how many files you can store [15:00]
gac410The issue will probably depend upon the underlying file system. Some are much better with large numbers of files than others. But I suppose for the 90% of installations, there really will not be 1000's of attachments. [15:00]
CDotand if the same attachment exists on two topics? [15:00]
jastthe next generation filesystems won't have any such limits [15:00]
MichaelDaumreuse it [15:00]
CDotnasty; why? [15:00]
MichaelDaumto make attachments first class citizens [15:01]
jastand fast access of files in huge directories is a non-issue these days [15:01]
CDot"the same attachment" may be two completely different files
I *could* encode the topic name into the attachment name; but if I do that, why not just use a directory?
[15:01]
MichaelDaumthis was actually an idea of Sven [15:02]
CDotand since the ,pfv has to exist anyway.... [15:02]
gac410jast ... and we are concerned with moving off of Perl 5.8.4 to 5.8.8 due to luddite (aka stable) distributions. Next generation file systems .... HA [15:02]
CDotmaking attachments first class implies a *lot* of rewriting of the UI etc [15:02]
jastgac410: I'm just stringing random words together here. :} [15:02]
CDotso.... I'm open to the idea, but it's way beyond the scope of what I'm trying to do here [15:02]
MichaelDaumSven was envisioning a store where everything is a meta object (topics, attachments, my-event meta data) [15:03]
CDotyeah, that's an old dream (/me remembers discussing it with MichaelSparks)
it would make life so much easier in so many ways
anyway, nice idea, but way beyond where I can get to, I'm afraid
[15:03]
MichaelDaummaybe not. it very much sounds like what git does with its object store [15:04]
CDotCDot is trying to implement the existing Foswiki::Store API as "sensibly" as he can [15:04]
FoswikiBothttp://trunk.foswiki.org/System/PerlDoc?module=Foswiki::Store [15:04]
jastit might be worth keeping in mind that for very, very big installations, people might want to offload attachments to a separate fileserver and have seamless integration of that [15:04]
CDotjast: that's a good point [15:05]
jastit's come up in one of our projects [15:05]
CDotbut I'm not at all sure we should be asking the PFS to do that [15:05]
MichaelDaumthe idea would be to store each revision into the file system and base its name on the checksum of the content ... [15:05]
CDotKISS has to be the principle [15:05]
jastprobably not :) [15:05]
gac410Hm. I have some stalled proposals to try to integrate Amazon S3, and other cloud storage for attachments. But I was approaching it as an event driven mirror. [15:05]
jastI had a fairly KISS idea for data deduplication, though [15:06]
gac410(I really want to offload my poor dsl uplink here :) ) [15:06]
jasthave a separate dir where all attachments are *actually* stored, with a name based on a hash on the contents
and the attachments in the "visible" locations are just symlinks
[15:06]
CDothow would you manage the de-duplication? [15:06]
jastupload the file, hash it, see that the hash already exists, discard uploaded file and use existing one instead [15:07]
MichaelDaumyea [15:07]
jastall that's left to do is an occasional cleanup of unreferenced hash files [15:07]
MichaelDaumthat could help multilingual content as well: by associating different objects to the same topic name ... though thats not possible with links and would need a proper mapping table instead [15:09]
gac410Ah. Still brainstorming, not a proposal: Foswiki:Development/EnableCloudStorageForAttachments [15:09]
FoswikiBothttp://foswiki.org/Development/EnableCloudStorageForAttachments [ EnableCloudStorageForAttachments ] [15:09]
gac410But it comes down to a mapping of the attachment to a storage location. [15:10]
MichaelDaumsame as for topics and whatever
it basically is simpler as hard-coding the web/topic/attachment anatomy into the structure of the storage engine
might even suffice to only hash the topicinfo instead of the full content
[15:10]
gac410Hashing the content would prevent storage of duplicate attachments. but would be more costly. [15:16]
MichaelDaumthough git is pretty fast in even hashing the full content...not sure whether there are shortcuts. [15:16]
gac410Maybe native i/o vs. perl would again be a problem. md5sum or sha1sum are pretty fast. But if perl did the calculation, ... [15:17]
MichaelDaumeven meta records holding the rev info are store objects in git [15:18]
CDothashing could result in the wrong attachments. Hash collisions are not unknown. [15:24]
gac410I always wondered about that ... how does git manage? [15:25]
MichaelDaumit uses strong enuf hash algos [15:26]
CDot^strong^long. See http://seejeffrun.blogspot.co.uk/2009/08/hash-collisions-in-git.html for an explanation.
I did consider using hash names in PFS, but rejected it for two reasons; first, it makes the datadir very obscure, and second, the time taken to compute the hash
seems to me it would make more sense to XS to a git library in that case :-)
[15:26]
gac410On the cloud storage, my assumption was that - all attachments would still be stored locally, the upload to the cloud might be costly. The cloud being more of a distributed cache than a primary store. [15:30]
Do we have any policy about foswiki.org registration using real names? AaaaBbbb and FirstNameLastName are stretching things a bit.
RebeccaUnger seems to possibly be a link farming page?
[15:35]
CDotno; we deliberately elected *not* to have such a policy. [15:45]
gac410okay.
How about users that register just to post links ... I had been treating them as spammers, even if they are only spamming their own user topic.
[15:46]
jastletting those kinds of users live is a direct contribution to junk results in search engines. just saying. :} [15:49]
CDoteven user topics are community property, IMHO. If someone is abusing the community by posting crap on their home topics, then the community has every right to take action.
the whole point of a wiki is that anyone can edit, after all!
[15:59]
jastproblem is that it's still in the revision history and so still improves the search ranking of the links in question [16:04]
CDotI thought we excluded robots from following links into the revision history.....?
assuming the robots respect nofollow, that is
[16:06]
jastI don't know whether that's in place
I suppose I'll just shut up, then. :)
[16:07]
gac410From what I can tell from the server logs, the robots follow every link including the nofollow links. IIRC google even says that they follow the links but don
don't index them. Which from a server load management issue is useless.
They also follow every table sort link in every table in every direction
[16:16]
***ChanServ sets mode: +o MichaelDaum_ [16:17]
gac410we would be better to actually not render a lot of the links for guests. [16:17]
jastgac410: I think the idea is that they follow the links but don't consider them for the ranking
not really bright but w/e
gotta go. enjoy your respective days/evenings/nights!
[16:19]
........................ (idle for 1h55mn)
***ChanServ sets mode: +o MichaelDaum [18:15]
.............................................................. (idle for 5h5mn)
chriswerry has left [23:20]

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