XML Blog Stuff IRC Ramblings

The Beginning

<nlindblad> anyone using LightPress then?
<nlindblad> http://lightpress.org/
<nlindblad> looks interesting
<sward> nlindblad: It's PHP (nevermind a plugin for another PHP app), I
        lost any interest I may have had in it
<nlindblad> sward: what CMS would you use then?
<sward> nlindblad: I don't :)
<nlindblad> sward: okey...
<nlindblad> sward: what would you recommend then
<sward> nlindblad: I wouldn't, I'm not a big fan of CMSes, at least not
        in the "dynamic page generation from static content in a data
        store which only looks dynamic because it's updated often" sense
<nlindblad> ah
<nlindblad> sward: true
<nlindblad> sward: I'm really after something I can update without
            having to use vim all the time
<sward> nlindblad: and vim's a problem because?  I find it better than a
        textarea
<nlindblad> sward: hmm
<h|barbobot> vim is no problem.
<nlindblad> sward: I'll just do a neat page then and use vim and shut
            up, etc.
<sward> nlindblad: Not that I have a website to manage.. I mean, the one
        page I have up hasn't changed for 3 years, except for updates to
        my CV and text ever so slightly modified to be past tense (I am
        studying.. -.  I studied, blah)
<nlindblad> :P
<StarScream> i really should make a website for my self some day
<sward> nlindblad: For anything more complex, I'd probably look into an
        XML-based system, but the resulting pages are generated by me
        (or automagically as I add content) (probably XSL
        transformations), rather than having the content pulled from a
        DB and generating pages dynamically.
<Dave2> XML!
<sward> nlindblad: Even with caching, dynamic generation in that way
        seems a bad method
<nlindblad> sward: TBH, Gentoo's XML policies are lovely
<nlindblad> sward: it's so easy to write content for their site
<nlindblad> sward: XML is neat
<nlindblad> sward: but the question is then
<nlindblad> sward: how would one go on about doing it properly
<nlindblad> sward: my idea is to make separate XML-documents for each
            "post" then merge them into a static page using Python in
            chronological order after their <date> tags
<sward> nlindblad: For journal type stuff, yeah, possibly.  You may not
        even need XML documents for each post, just plain text which
        gets included into XML, though that doesn't help with how you
        get the date.  I don't think file modification times are
        sufficient.
<nlindblad> yeah
<nlindblad> http://pastebin.com/683246
<sward> Heh, just had a thought.. changelog format is essentially the
        same as a journal, and vim has nice macros to manipulate
        changelogs
<h|barbobot> vim uber alles!
<sward> nlindblad: In your example post, too much crap.  You want it
        simple.
<nlindblad> well
<sward> nlindblad: It can be an XML fragment, rather than a complete XML
        document, so you can probably leave out the header crap
<nlindblad> <time> is probably crap
<nlindblad> sward: true
<nlindblad> http://pastebin.com/683252
<sward> nlindblad: Yeah, make date + time = <date>, and in ISP format
<sward> ISO
<nlindblad> yeah
<sward> nlindblad: If you want to separate them, it's designed to be
        easy to process, so... process it :)
<nlindblad> http://pastebin.com/683255 then
<nlindblad> the point of having <title> is for automatically generating
            the navigation
<nlindblad> and to have a real title that is a bit bigger than the
            content
<nlindblad> sward: the idea is that you shouldn't have to care about
            what you name the files
<sward> Agreed
<nlindblad> sward: as long as you put all the posts in a pre-defined
            directory the Python-script should look at the date and
            properly generate
<sward> although I think nice names are better than some crap hash
<nlindblad> yeah
<nlindblad> the ideal thing would be to have a vim-skeleton defined for
            the post directory
<nlindblad> so vim /postdirectory/somefile
<nlindblad> would create a new skeleton with empty tags
<nlindblad> (if somefile doesn't exist)
<sward> also, if your blog links to "full" versions of posts where you
        can add comments and such, then the server should serve it as
        http://host.domain.tld/blog/nice_name instead of crap like
        http://host.domain.tld/blog/2006/04/21/<number-assigned-to-the-current-phase-of-the-moons-of-jupiter>/$(cat
        /dev/urandom | base64encoder)/shit_name
<nlindblad> sward: hmm
<sward> date, long title, author.. all meta-data rather than URI
        material, though that does bring the "how do you ensure each
        name is unique?" question
<nlindblad> sward: I don't intend to keep the directory with posts in
            the htdocs
<nlindblad> sward: so what you're saying makes perfect sense, generate
            standalone pages and make nice URIs
<sward> I wouldn't, but you want to be able to generate reasonable
        output files
<nlindblad> sward: what tags would you use then?
<nlindblad> to be exact
<sward> not sure.. within the confines of XML, you've about minimised
        the amount of faff you need to put in as much as you can
<sward> you'd probably do better generating an index
<astopy> putting the date in the URI is a good idea
<sward> then you can setup something to "add a post", and it'll fill in
        the date for you, so you can get rid of that
<nlindblad> sward: yeah, preferably a small Python-script simply taking
            input
<sward> once the post is in the index file.. it won't be overwritten
        (unless you tell "add a post" to), so existing dates won't get
        changed to the current date
<nlindblad> sward: but somewhere along the road I think I should be a
            bit different to other traditional blogs
<sward> I know someone who did a set of XML templates for his site at
        one point, maybe I'll steal from him
<nlindblad> http://pastebin.com/683281
<nlindblad> yes, alot of tags
<nlindblad> but the metadata helps you create some cool stuff out of it
<sward> it might be useful to separate metadata from the content.. most
        of that I'd want to auto generate
<nlindblad> sward: you mean to put "easy to generate" stuff in a
            <metadata> tag?
<sward> for example, if I wanted to create an id (reference) for a
        post.. I'd process the title into some canonical form which is
        valid for ids
<nlindblad> true
<nlindblad> the thing is
<sward> name can be obtained from your user name, e.g. like mutt fills
        in 'Simon Edward Ward' in my From: address
<nlindblad> I must settle for a DTD before I can continue doing real
            code
<nlindblad> yeah
<nlindblad> http://pastebin.com/683287
<nlindblad> now all data intended to be auto-added is in the <meta> tag
<nlindblad> *are
<sward> nlindblad: I'll start putting stuff up at
        http://projects.bleah.co.uk/misc/wiki/XmlBlog .. I'm actually
        interested in seeing this work
<sward> nlindblad: Feel free to edit the page and add new sub-pages
        (assuming I let anonymous do that...)
<nlindblad> okey
<nlindblad> I will
<nlindblad> I'll dig through my XML books later on
<nlindblad> but I think http://pastebin.com/683304 is a good start
<nlindblad> if I've understood DTD a bit
<sward> balls, the person who I thought did a pure XML-based template
        system didn't.. he used Smarty
<nlindblad> :(
<nlindblad> sward: but that is all the DTD required to syntax-check a
            post
<sward> It should be easy enough, given this XML system, to write stuff
        to post using the blogger API too
<nlindblad> sward: yeah
<sward> XSLT transform to blogger API XML-RPC... HTTP POST
<sward> well, probably not that simple, but it was an idea :)
<nlindblad> when the basic XML is completed and the parsing is also done
<nlindblad> extending will be very easy
* nlindblad wonders whether a XML-validator in Python would require
   hours and hours of work
<resiak> I'm sure that one exists
<sward> if not, the various XML libs are a good way there
<nlindblad> I've done some simple thing with SAX
* sward -> pub -> home
<nlindblad> http://pastebin.com/683364
<nlindblad> where to go on from here?
<nlindblad> the idea is to make the parsed data ready for output