Re: Bookmark 'stability'
"Eric J. Bowman" <[email protected]>
| Newsgroups | gmane.comp.web.services.rest |
|---|---|
| Organization | Bison Systems Corporation |
| Message-ID | <[email protected]> |
Mike Kelly wrote: > > Hence why we have 'entry points' which /are/ the stable URIs of a > hypermedia application. This is what distinguishes them as an entry > point - they are an resource a client can assume exists (i.e. > bookmark). > This notion of 'entry points' is a common misconception about REST. The 'bookmark' relation is not a declaration of status. Every representation in a REST application is a potential entry point, that's what the hypertext constraint is all about -- there's no set application flow. Think of a shopping cart app -- I may go back and change the ship-to address at a later point than I was initially asked, whereas you did not. I may have entered from a product page, whereas you may have entered from a special-offer ad. Does the fact that we used different starting points and followed a different path through the process, mean we used separate shopping-cart apps? Roy's whole point with bookmarking is to contrast with the RPC style, where you can't go back and edit a ship-to address, you can only go back to step one and repeat the whole process. I may bookmark the "payment" page because the wife's out shopping with the credit card, so I want to come back later and *that* is my entry point. > > If every URI in your app was supposed to be bookmarkable then the > term 'entry point' would be meaningless. > But it *is* meaningless to a RESTful system, which is my whole point in this thread. -Eric