Re: REST descriptive services (was Re: [rest-discuss] REST description using WSDL)
- Hi Roy,
On Nov 3, 2005, at 9:46 AM, Roy T. Fielding wrote:
> On Nov 3, 2005, at 8:47 AM, Nic Ferrier wrote:
> > One disadvantage to REST is that it can't be auto-discovered and
> > turned into an API for a traditional language in quite the same
> way as
> > the WSDL/SOAP stack.
> You mean REST-based HTTP, right? We don't need to "auto-discover"
> an HTTP interface because the communication API is a standard and
> the parameters are defined by the current representation. We do
> need to discover the initial resource, and that's exactly what
> Yahoo, Google, and URIs written on napkins do for us. Any well-known
> website for "finding services" would do the same.
I know that's true in theory, but I still don't know how to implement
that in practice.
For example, say that Amazon had a truly RESTful API. I wanted to
build an automatically-generated GUI client that would be able to
query, browse, and display anything accessible from their datastore.
Given just the original URL, how does my automated tool:
a) Figure out how to make queries
b) Discover which disjoint URI spaces might exist
c) Unpack the different fields and datatypes from results
For that matter, if I'm writing a REST web services, what information
do I need to provide -- and where -- to enable such a tool?
Maybe this is obvious to you, but it sure isn't to me, so I'd like to
have something more actionable.
Or are you merely saying that such tools are not interesting?
-- Ernie P.
> > Ugh.I am responsible for the primary storage of offers, prices and inventory for
> > I'll make sure stuff in my area doesn't do it this way.
> > Next time I talk with a person from the cart group, I'll
> bug them about it.
> What's your area, and how will you create and modify
> resources (in contrast to CartModify)?
all listings on the Amazon platform, worldwide.
Well, almost... I'm not responsible for inventory that comes from Amazon
fulfillment centers. And a large portion of Amazon offer information
originates from fairly rich systems that send data to my services after
running 'their' business rules (which products to offer, at what price,
My system is mainly merchant-facing and not web-facing - data is published
out of my system into other 'caching layers'.
There are many teams that provide many services, so this isn't a complete
story by any means.
But I'm hiring...
The data is submitted by external sellers and merchants through bulk feeds
and through browser-based tools. The "Sell Yours Here" button is an example
of a browser based tool. (My team doesn't do the UI, but does own the
business logic and accept/store/publish the listings). There are two
'platforms' - Marketplace and Merchants. The Marketplace platform is open
for registration here:
http://www.amazon.com/exec/obidos/subst/misc/sell-your-stuff.html and paying
for a 'Pro Merchant' account gives access to bulk uploads. There are also
third-party tools and services that do this integration for you. The
Merchants platform is invite-only (at this time) and supports more complex
feeds that allow for creating new products in the Amazon master catalog in
bulk (these result in new details pages - ASINs) beyond just listing what is
>I don't know. I'd rather not guess out loud until I talk with them.
> And why do you think they did it that way? (Assuming it was
> a reasoned design decision...)