I think the use case approach is the best way to unpack this. As such
it seems we have been discussing three use cases:
1 Display a node and its properties
2 Display a node, its properties, and its children
3 Display a node its properties, and the full structure underneath it
If those are the use cases then I think the first one is the most
basic case and should be the default, the other two should require
some sort of addition to the URI. On the other hand, I guess, the
discussion boils down to how we perceive the json request. Are we
asking to serialize an object graph? If so then we should dump the
whole thing. I am not sure, however, if I am comfortable with that
given the potential consequences of dumping tons of data.
Going back the the filesystem analogy given this structure:
/
/foo
/foo/child1
/bar
/bar/child2
When I ask to list / I do not get a recursive listing unless I
explicitly ask for it. Using the use cases above how would we expect
request calls to work? What should the URIs look like? I think I agree
with Toby that UC1 and UC2 should be collapsed so that a request for /
foo would return foo's attributes plus list foo's child node names. A
request for /foo.recurse would return /foo + all its children their
properties their children, and so on... With those two types of
request one could fulfill the browsing and the dump models Toby
described. I also agree that these should be done with selectors and
not query params as it makes caching much easier.
-paddy
On Jan 9, 2008, at 1:01 PM, Tobias Bocanegra wrote:
Why not:
/stuff.json?depth=1
It's not like we need to use the URI mapping for everything.
because this breaks caching.
however, i think there are actually 2 use cases. the first is that you
need the deep serialization of a tree, e.g. a page, or an exploded
xml, or a dialog node structure description.... in this case you
always want a depth=infinity. the other one is the 'browsing' use case
where you only need the node and it's properties (and the names of the
childnodes!). so this is a special case anyways. this would be like
depth=0.5 :-)
so is there really a use case for depth=0 (i.e. node+properties) ?
or is depth=0 just the node name, depth=1 the node name and it's
properties and it's childnode names? etc. ?
regards, toby
--
-----------------------------------------< [EMAIL PROTECTED]
>---
Tobias Bocanegra, Day Management AG, Barfuesserplatz 6, CH - 4001
Basel
T +41 61 226 98 98, F +41 61 226 98 97
-----------------------------------------------< http://www.day.com
>---
