The original advogato.org engine, mod_virgule, was first released in November 1999. One of the things that Raph mentioned he wanted was less hard-coded c, more XML. Well, virgule resulted, from four months hard-core code morphing, as a fledgeling forum scripting language. Development of this project, after a six month wait, is now underway again, at http://virgule.sourceforge.net.
Well, i figured that adding usage of APR (http://apr.apache.org) was worthwhile enough to consider the development of the xvl engine "active", especially as it was possible to re-enable the cgi-bin capability again, as a result.
One of the things missing from mod_virgule is the capability to run external executables, such as automatically emailing users passwords to them, running cvs, news, etc., all of which are on the feature 'wish-list' for more advanced virgule forums.
Well, they're now feasible.
I'm in the process of getting a development site set up that _uses_ xvl [nothing like believing in your own technology, neh? :) ] so will follow-up later and announce when that happens.
This implementation has been accused of being an all-singing, all-dancing scripting engine. If _that's_ the case, then php is the fat lady queen of the opera _and_ you have to go to her concerts every day. uggh :)
It can't claim to be all-singing all-dancing [only php can]. Fascinating? yes. Suffering from first-stage design / implementation warts? yes. Stable? yes. Useable? yes. XML-data-driven? yes. Not for the faint-hearted? Yes.
I like it :)
Could you enumerate the requirements for a web scripting engine to be all-singing, all-dancing?
hm. let's see.
you'd need... an interface to the sound device. html client to locate mp3s... a script to locate the shutdown command and remove it... a small amount of assembler to access the bios and disable power-off mode... i'm not sure how you'd arrange to stop people pulling the plug out the back of the machine, though.
i'll leave the all-dancing one to your imagination, shall i? :)
Arguably, you want a high level language for this stuff and arguably you are defining one.
Why not write the whole thing in Perl, or Python, or Lisp, or ...?
hiya,
well, i ported the trust metric code to python, already. see pymmetry - python trust metrics. however, that doesn't make a community-based web server :)
the answer to your question is, i don't really know! maybe it's because i want to see how far i can push this concept, it's not exactly a new one: there already exists an XML scripting language, called xsl. i borrowed the o'reilly xsl in a nutshell book, read it, and was impressed by bits of it and not by others (for example, its boolean expression/evaluations are put into a property as if xsl was a standard scripting language) whereas xvl leverages tags, which makes for a much simpler code-base, by having <expr:if>, <expr:then>, <expr:else> and <expr:eval>. you get and, or and not, and full if-then-else syntax, and this is implemented in a grand total of 326 lines of c code. for c-code, that's almost unheard of, these days :)
xvl is like, dedicated to XML data manipulation, and feeding that XML data to and from HTML. with those simple principles, and the extensibility via dynamic module loading, doing whatever-you-like is actually pretty trivial.
the whole point of python is 'Do it Once, Do It Right'.
the whole point of perl is 'Do It Many Times' [afaict! :)]
so python is slim, and yet highly advanced and functional. xvl is slim to the point of being a beanstalk, at 12-15,000 LOC, and still functional enough to do a feature-rich web site [not that i can write good HTML to fully demonstrate that myself, however that shouldn't stop anyone else! :) :)]
so, truthful answer to your question: i don't know! and until i find out, i'm going to carry on, because i can, this is open source, and i hope that APR and other projects will benefit as an indirect result :) :)
what xslt is designed for is to do exactly what xvl does, too: take in XML documents, spit out XML documents. i think it's probably worthwhile investigating. i'd love to not have to write an entire set of new data manipulation constructs and to use a pre-existing feature-complete codebase.
i wonder how it may be possible to add a <tm:calc /> tag so that i can do the trust metric calculations off of the user-certifications?
i wonder, also, if the xslt team have plans to write an apache module, to serve HTML from xml templates?
curious and interesting.
It's interesting that Lions thought that 9-10k loc was the limit that his students and most programmers could be expected to understand, which is why he liked using Unix v6 for teaching as the kernel was that small.
It's taken some playing but I've got a test website running bits and pieces of this, and so far, I can see the great potential of it.
As you said, Luke, not for the faint of heart, but with just a little mangling, I did get it running. Something is wrong with either the XML schema, aprs, or the c code, because the options tags didn't end up as valid HTML select options. Hacking the c code to make it write good tags fixed it for me, and as I add sample data to the site, most of the pages are starting to fill out correctly, since they tend to die/look odd without a single database entry in many cases.
The cert stuff is still badly confused (it will let me Certify things as people and vice versa) and there are other bugs I'm noticing as I play with it.
Features that are in this that Advogato lacks (for those interested):
Messaging - how many times did you read something in a diary and want to write a reply to it you are sure the other person will see?
News and Articles - Articles same as Advogato's but News needs to be certed before it gets posted. Articles can be created under a project too.
Tasks - per person or per project, task list creation.
URLs - create a list and even view targets with a frame. (bug: I can only seem to create one URL and edit it, not create a 2nd one...)
Certifying - can list reason for rating, multiple kinds of ratings, including the main one, staff, project interest, project connection. Extendable pretty easily. (In fact, that's why I'm looking at this, I want to do a discussion website completely un-software related, needing trust metrics, but want to remove all of the project/master/developer references and replace with something similar but not just a name switch (ala badvogato), and still easy enough that computer newbies can navigate it and still participate in trust issues/ratings.
overall, very rough and not yet ready for someone to plug and play, but clearly a logical step up from the current codebase and worth the wait!
thx seth.
yes, you are allowed to certify projects, articles, news and urls, as well as people. in fact, in the xmlvl example, the only way to get news onto the frontpage is for it to have a Certification of Apprentice or above. no cert, no trusted, no news.
also, as a test, try Certifying a test project as 'Interesting' and then go to your /person/username page: you will see that the cert is listed under a heading 'This person has the following Interests'.
also, if you Certify a person as 'None', meaning 'No Interest', then no-one, including you, will ever see messages sent to you from that person on your /person/username page. so, if someone starts spamming you, you just Certify them as 'None', and their offensive messages disappear [the default Certification level for Interest is 'Ambivalent', btw].
this is just a couple of examples of the uses of advanced Certification: it effectively becomes like a Search Engine, as thom and others have pointed out, in other articles about xvl.
... *click*. did you use the advogato.org example site or the xmlvl example site?
i haven't updated the advogato.org example for quite a while, my main focus at the mo is on the xmlvl one to get it ready.
i've added tasks to people and projects, and am working on a way to do dependencies / task management.
and discussion (reply) to projects and urls. i considered adding replies to diary entries, but it requires a bit of thought/redesign: diary entries are *already* a sub-directoried entity, and adding subdirs to subdirs will require a little bit of redesign.
oh, and much as people might like to, i'm _not_ going to add discussion threads to people - it _is_ supposed to be a public forum, after all ;)
luke
You are allowed to certify projects, articles, news and urls, as well as people.
I understand that, but (at least in my test site) it is offering me _all_ of the choices for everything. I can see certing a person as Interesting, Master, Documenter, etc, but a project as Master? That's confusing.
in fact, in the xmlvl example, the only way to get news onto the frontpage is for it to have a Certification of Apprentice or above. no cert, no trusted, no news.
Yes, I noticed and liked that, but Certing news as Apprentice doesn't make sense to me, using the Interest cert instead does. It would nice if the different cert schemas has limits as to what can apply to what... I thought they did, till I played with it. It offers you all of the different certs for everything.
also, if you Certify a person as 'None', meaning 'No Interest', then no- one, including you, will ever see messages sent to you from that person on your /person/username page. so, if someone starts spamming you, you just Certify them as 'None', and their offensive messages disappear [the default Certification level for Interest is 'Ambivalent', btw].
ah ha, I was wondering about that... the documentation is still so sketchy. Maybe a negative cert would be a better filter name. None still seems neutral to me. If Interesting is a good cert, maybe Offensive would be a zero/bad cert? So that if I found a topic/person/url I didn't like, rate it offensive and it's gone... at least for my view...
...as thom and others have pointed out, in other articles about xvl.
Where are those articles? I can't find them. I posted on the xmlvl sourceforge list and got no answer, and there is (almost) nothing in the archive, so I decided posting here would at least get some notice (grin)
... *click*. did you use the advogato.org example site or the xmlvl example site?
the xmlvl site... looked far more complete to me. It's still not a good general sample site, though, so I've been working on that, trying to minimize the specific stuff and make it look as general as possible, so I can figure out what I want to keep and what I want to change (and where to make those changes) The best way I've found to figure what something this complex does is to try reducing it down until it doesn't work, then restore it bit by bit.
Yes, I've seen this, or an earlier version also by lkcl in action at the open-it web page.
It's very neat, very very neat.
well, thanks to DV's input, from his diary entry 2day ago, i'm adding support for <xincludes/> and will also add in xslt - XML Schema Language Templating.
the combination of xvl and xslt is, well... the possibilities are very powerful (not least because xslt is a recognised standard)
xslt is a tool that can be used _by_ XML-based databases. it is mandatory to specify two files as input, one the source doc, the other the template. this makes it impossible to use xslt to drive, say, a web site, unless you... oh, _i_ get it.
oh i LIKE it. of course! _that_'s why xslt is being used to generate WAP site content.
of course, ha ha! you can use xslt to template-drive a site for the format, and the 'data' is the site's content!
so, you have one template for a WAP client, one template for a netscape client, one template for lynx, another for internet destroyer, another for no-frames, another for text-only, and... say... another with or without style sheets! and by using xslt to 'merge' the template and the site content, they all get the same data with minimum site-development effort.
hey, that's so wicked!
I understand that, but (at least in my test site) it is offering me _all_ of the choices for everything. I can see certing a person as Interesting, Master, Documenter, etc, but a project as Master? That's confusing.
okay. well, there are two reasons for being able to Cert a project as Master. 1) ... you can potentially login as a project. i haven't enabled this, but it _is_ possible. 2) i intend for 'Project Developers' and 'Project Documenters' to be able to gain privileges *by way of the project having been certified by real people*.
imagine a weighted graph, like the advogato cert-net. now add projects to it, not just people. _now_ add 'developers' and 'documenters' as edges of a special type in that graph.
what do you have [other than a mess] when you do Trust-Metric evaluation on that? an information-loaded hierarchy representing people's roles in projects, and a way to assign them rights to, say, cvs commit, manage a project, etc.
in fact, in the xmlvl example, the only way to get news onto the frontpage is for it to have a Certification of Apprentice or above. no cert, no trusted, no news.
Yes, I noticed and liked that, but Certing news as Apprentice doesn't make sense to me, using the Interest cert instead does.okay, well, it kinda-sorta-makes-sense to me, but i do know what you mean. designing Certs is tricky [aside from having to get the grammatical syntax right, you have to pick either all-nouns, or all-adverbs, or all-adjectives yick].
if you can think of something, please - do let me know, and i will create it. see, the purpose of applying 'Master' to news is that if someone deems an item to be news-worthy and within the Charter of the site, they Certify it. it then, if they have Weight, will go onto the front-page.
Interest doesn't really cut it, alone. Interest was designed to be of 'Personal' interest, for use a bit like a 'number-of-degrees-from' search-engine.
however, you are right: it may be good to 'customise' a site by displaying, on the front-page, things of 'Interest' to you.
but, as you can see, that's not really necessary. you go to someone's home-page, you see they have Certified a news article as 'Interesting', you click on it. or, you go to your _own_ home page, you see things that _you_ have Certed as Interesting, you click on _them_.
couldn't be simpler, really.
btw i haven't implemented 'double-level' Trust Metric evaluations [i.e. into the netflow code you put *two* certs not just one like there is at the moment] by type-of-cert, yet, but i _have_ implemented a means to 'multiply' two Certs together, see <tm:chart />, to produce 'ratings', which i totally hate.
It would nice if the different cert schemas has limits as to what can apply to what... I thought they did, till I played with it.
yes, they do, and i applied them to lots of things, but not all-to-all. e.g. i don't apply the 'staff.xml' cert to news or articles.
see the certs/ directory, for example, see the certs/interest.xml you will find that a) it has no seeds (because the Interest Cert is relative to individual tastes, not to the tastes of the site maintainer) b) it applies to proj, news, article, person, url.
the certs/staff.xml cert applies to people and proj, and again, has no seeds because the seed is *automatically* taken as the project to which the Developers belong. anyone who has registered that they are a Developer for someone else will *automatically* be listed under the Project Staff, via their... errr... Manager. whatever.
It offers you all of the different certs for everything.
not quite.
also, if you Certify a person as 'None', meaning 'No Interest', then no- one, including you, will ever see messages sent to you from that person on your /person/username page. so, if someone starts spamming you, you just Certify them as 'None', and their offensive messages disappear [the default Certification level for Interest is 'Ambivalent', btw].ah ha, I was wondering about that... the documentation is still so sketchy.
sorry :)
Maybe a negative cert would be a better filter name. None still seems neutral to me. If Interesting is a good cert, maybe Offensive would be a zero/bad cert? So that if I found a topic/person/url I didn't like, rate it offensive and it's gone... at least for my view...
i'm giving a talk on trust metrics and social engineering at the ukuug in manchester, end of this month. one of the reasons why you _don't_ add Negative Certs is because it then encourages idiots to use it for their own personal amusement. remember the Dummy Rating, and how much fun it was for people with no other authority on the site to Certify rms as Dummy?
_that_ doesn't sound like it's in the Charter of the advogato.org site [to promote open source dev / discussion]!
...as thom and others have pointed out, in other articles about xvl.Where are those articles? I can't find them. I posted on the xmlvl sourceforge list and got no answer, and there is (almost) nothing in the archive, so I decided posting here would at least get some notice (grin)
eek! you did??? hardly anyone uses it!!! :) :) maybe i should switch it off or something.
... *click*. did you use the advogato.org example site or the xmlvl example site?the xmlvl site... looked far more complete to me. It's still not a good general sample site, though, so I've been working on that, trying to minimize the specific stuff and make it look as general as possible, so I can figure out what I want to keep and what I want to change (and where to make those changes) The best way I've found to figure what something this complex does is to try reducing it down until it doesn't work, then restore it bit by bit.
one of the things i wanted to do was to create a 'basic', simple site. if you'd like to help do that, i'd be honoured. my focus at the mo is on getting xmlvl ready for launch when the server goes live, however, the whole point of _that_ exercise is to find other people who may be able to tell me if this is a Good Thing or not. so, if you're willing to explore, i'm willing to help clear the path.
luke.
thanks nigel.
well, _if_ advogato was running an 'Interests' Cert, or had the URL capability, i could point you at my homepage and it would be easy: i would have Certified all my own articles as 'Interesting'.
:) :)
when xmlvl goes live, i will trawl through the internet and create lots of links and references etc.
[hm, this spells.... i have an idea. i should 'categorise' URLs, by a group-name. hmmm. :) at the moment, they are just numbered, like articles. hmmm....]
wha-heey! the simple tests work. i moved on to a larger one (the XML spec) and that bombed out half-way through [at the xvl end, not libxslt]. not bad for 2 hours quick-hacking: will investigate over w/e and find out wassup.
well, it was incredibly easy. a simple <schema:table /> get and i have the user's password, for mod_auth_xvl, and a simple <cert:check /> and i have mod_access_xvl. cut/paste from mod_auth.c and mod_acccess.c, and it's done.
v. impressed with apache.
in case i haven't mentioned it here, yet, i also morphed the xvl codebase so that .xvl is now a registered mime-type.
one of the serious limitations of mod_virgule is that it 'takes over' an entire directory, controlling the entire site from there. whilst i originally thought this was cool, i decided that a more 'conventional' approach would have far more benefits. such as not having to have mod_virgule do its *own* loading and decoding of mpeg, jpg and the scores of other mime types that apache handles.
with mod_virgule, mixing php in the same directory is _totally_ out of the question. with mod_xvl / xvl, you can.
New HTML Parser: The long-awaited libxml2 based HTML parser code is live. It needs further work but already handles most markup better than the original parser.
Keep up with the latest Advogato features by reading the Advogato status blog.
If you're a C programmer with some spare time, take a look at the mod_virgule project page and help us with one of the tasks on the ToDo list!
del.icio.us
Digg
Google bookmark
reddit
Simpy
StumbleUpon
Furl
Newsvine
Technorati
Tailrank