Location Resolution
-----------------------
These are notes from a discussion on using URN-type names instead of filenames for locationmap
resolution. This is a summary of the the relevant pieces of the IRC discussion below.
As a part of the making deeper use of Locationmaps in the core Sitemaps with a goal of
letting locationmaps do all location resolution and letting sitemaps stick to the pipeline
processing. For those that are familiar with name resolution servers or the Handles Service,
it might be easier to think of the locationmap as a name resolution module or sort of a
handle resolution module in that it accepts "names" or whatever you desire to call these
"hints" and returns the location. The thought is that by using file name "looking" hints
it disguises what locationmaps are really doing for us. By using URN-style names, we are
truly disassociating the name/hint form the physical location.
For example, here is a locationmap entry based purely on filename:
and now below is that same entry using a "name" style. One implies a certain physical location
where as the one below is truly a name that needs to be resolved to a physical location.
The format is essentially:
resource-type(dot)transform-type(dot)from-format(dot)to-format
resource-type(dot)type(dot)name
Examples of these two:
transform.xslt.xthml2.html
graphic.png.project-logo
twilliams_ i've been thinking lately about using URN's for locationmap resolution to totally separate the physicall locations. I realize you're wanting to get stuff done but maybe just think about it
* rgardler has quit (Remote closed the connection)
* rgardler (n=rgardler@ip-213-92-134-135.aramiska-arc.aramiska.net) has joined #for-s
twilliams_ was wondering if you broke it and took the rest of the night of;)
rgardler No, it was the wireless going down again, so I took a wlk to the sea. Bacl for another hour (I only paid for wireless until midnight GMT)
rgardler So, you've been thinking of URN's for locationmap resolving - tell me more
twilliams_ did your latest commit's work on your machine?
rgardler Yes (I think ;-)
twilliams_ for some sources, it's necessary to pass the path as matched in; for others like the ones we'll find ourselves doing more of in the near future will be for named "things"
rgardler OK
twilliams_ i'm just thinking it' look cleaner if the named things weren't {lm:mystylesheet.xsl} but were a standard scheme of some sort
rgardler Sounds good...
twilliams_ true urns won't work I think because of the ":"
rgardler (about the plugin - I have broken the index.xml page because it is not an XHTML2 page, but the sample works)
rgardler I'm listening regarding the URNs idea...
twilliams_ but we could do "." instead (e.g. {lm:graphic.project.logo} or something
twilliams_ okay, i was only trying the index page
twilliams_ i didn't see .xml on the lm matches though
twilliams_ e.g.
rgardler It's there (straight after the comment
twilliams_ I've got: (no .xml) am i missing something?
twilliams_ the xml isn't included in {0}
rgardler {0} means all of the pattern (i.e. including the .xml)
twilliams_ ok i goofed locally
twilliams_ essentially, instead of having a bunch of filenames in {lm:}'s it'd be some named resources. Stylesheets aren't so clean though
rgardler I just turned the index page into an XHTML page so it works as well now (svn up)
rgardler So, tell me more about the URN idea
rgardler What advantage does it give us over a defined URL space?
twilliams_ looks better and not associated with a physical location (ie filename)
twilliams_ may be cosmetic i suppose but it makes sense that we're going through the work of separating locations from pipeline processing might as well take it one step further
rgardler Ok, I don't see it as a filename, rather a hint
twilliams_ stylesheets don't fit very well obviously
rgardler however, I agree it *looks* like a filename
twilliams_ technically it is a hint
twilliams_ but, when a user sees xthml2html.xsl -- they don't think hint
rgardler so lets make it *not* look like a filename
twilliams_ that's essentially what i'm saying
rgardler :-) Damned IRC - we have the same thoughts but can't communicate them fast enough
rgardler So, what would you suggest in place of xhtml2html.xl?
twilliams_ not sure, stylesheets present the problem because it's not a simple name, we're conveying a lot of meaning in there
rgardler I would be tempted to do something like
twilliams_ transform.xhtml2.html would work if documented that the syntax was resource-type(dot)from-format(dot)to-format
twilliams_ where resource-type could be: transform, style, graphic, etc.
rgardler Looks good (close to what I was going to suggest)...
rgardler The problem I see is that it does not indicate what type of transformer it is...
rgardler i.e. XSLT, chaperon etc...
rgardler We could add a "tranformation type" in there?
twilliams_ yeah, transform.xslt.xhtml.html that's almost long enough to become offensive;)
rgardler Yes, the length is a problem, but it does convey the meaning we need.