Re: Bookmark 'stability'
Paul Cohen <[email protected]>
| Newsgroups | gmane.comp.web.services.rest |
|---|---|
| Message-ID | <CAA1Y72F8gJYq_aE0jFhUiRFejQt5xkodUUYAPBz=YrUwmgas_g@mail.gmail.com> |
Hi, On Sat, Oct 6, 2012 at 10:11 PM, Eric J. Bowman <[email protected]> wrote: > Paul Cohen wrote: >> >> A server should commit itself to some "official URI:s", or entry >> points. But forcing a server to commit itself to maintaining all URI:s >> that have ever been published to any client seems to me like >> introducing unneccesary coupling. >> > > How is this coupling? I mean that *forcing* a server to commit itself to maintaining URI:s is coupling. I say that that the server has no such obligation. A polite server will issue redirects for old URI:s or maybe simply 404:s. I mean that having a requirement on the (developers of the) server to always maintain old URI:s, even though they'd rather just have the server return 404:s is an example of coupling. I am aware that there are many more aspects of coupling. > I take over a website for a client, change the > URI allocation scheme, then set up redirection on UltraDNS. Since REST > is a layered architecture, maintaining obsolete URIs doesn't even have > to be part of the new system. It's a courtesy to users, but even that > isn't a requirement. That was exactly my point. It is polite but not a requirement. The server-side application is free to do what it wants. > Installing a new system using new URIs is *not* a > REST violation, even if the old links die outright, the 404 response > can include a search interface or link to the new homepage, meeting the > hypertext constraint -- again, how is this coupling? It is coupling by virtue of *forcing* the server to maintain URI:s that the developers wish to remove. Of course, for practical purposes, any server that wishes to be polite and keep clients happy should take care to at least maintain some well known server URI:s. /Paul -- Paul Cohen www.seibostudios.se mobile: +46 730 787 035 e-mail: [email protected]