Hi Chris,

this sound quite promising. Interesting. Mind-widening. :-D

But as far as I understand REST and RESTful, I cannot quite map your description of Project: Yellowcake to the principles and terms I know. Please have a look at: http://en.wikipedia.org/wiki/REST

REST over HTTP means using the "Verbs" GET, PUT, DELETE, POST on resources identified by some kind of URL. But using PUT and DELETE in the Domino Web server?

Okay, this is only the usual interpretation of REST over HTTP, but not the only one possible.

What are your resource identifiers and verbs, than?

Thomas

I'm Looking forward to seeing more on this, but like Thomas, I'm seeing a conflict with my understanding of REST.

--
Kerr

Hi Chris,

I think, Project: Yellowcake can be something really new, enlightening and very powerful, but putting the "REST" label on it isn't necessary for its usefulness, but can provoke some "resistance" due to the unusual interpretation of "REST".

The key concepts of REST are resources and verbs. I don't see the verbs in your description nor example.

REST over HTTP is not only using GET and POST, but also PUT and DELETE (as they were intended and described in the HTT protocol). For example
DELETE /employees/1992
should delete the employee 1992 and
POST /employee/1993
with some parameters in the body create or modifiy the employee.
Perhaps, if you use different agents for the verbs and the first parameter as resource identifier...

Again, I don't say anything negative about your project, but I only recommend, that you rethink the unnecessary usage of the label "REST".

Cordially,
Thomas

I've been pressing F5 on your blog for two weeks, it was well worth waiting!
This looks very promising indeed.

Just don't go hanging out with those yellowbleeders to much :-)

Thomas, if your complaint is that it's not truly RESTful because it doesn't support PUT and DELETE commands, we can all blame IBM for not having a CGI handler that does anything with those commands.

I suppose Yellowcake could be re-written as a Java servlet, and then there's be something to PUT to, but Chris is certainly limited by which HTTP methods Domino will even acknowledge.

Would you guys object to "semi-RESTful?" What about "Sleepy?" "Dozing?" "Nappy?"

Would that mean that an application using Yellowcake would be Nappy-headed? :-D

For the record, I was the one that suggested to Chris that he call it a REST API when he said he was going to support POST as well as GET. So your beef is more with me than with Chris.

Hi Nathan, I don't have any meat with your nor Chris - no beef, no sheep, no chicken ;-)

BTW, the Domino URL commands could be called a REST API: ?OpenDocument, ?CreateDocument, ?SaveDocument and ?DeleteDocument are appropriate verbs and the path before the command identifies a resource properly.

Chris, the potentials I immediately see in your solution are the potent and flexible "Templating" approach, the additional abstration layer (dbkey, appkey), and the possibility for templates to use/modify data in more than one database concurrently.

Performance-wise a servlet could be beneficial and more flexible (thinking about PUT and DELETE, but also about the construction of resource identifiers aka URLs), but an agent could be "just good enough".

Writing about performance: Have you done some performance tests yet? This could get a weakness... Something to watch...

Waiting for more input :-)
Thomas

Again, for the record, I'm playing nice. Thomas and his colleague Bernd... I would dare to call them friends. Or at least brothers-in-arms. We fight for the common good of OOP in Lotuscript! :-)

So there's never any doubt: I have the utmost respect for assono.de!

I just want to deflect a bit of the purist approach to what you're offering here. Technically, a REST API for Domino is impossible from anyone besides IBM unless you want to build a DSAPI filter, simply because the CGI processor doesn't know what to do with PUT and DELETE methods.

And while DSAPI is vaguely cool from a technology standpoint, it is essentially a giant platform-specific hack -- and I'm quite sure that Thomas (nor Kerr) wouldn't find a DSAPI program a satisfying solution.

As to actual performance metrics, well, I got to smoke test some queries. They're fast. And not to promote other technologies, but honestly -- Yellowcake follows in the same vein as both Ext.ND and Decepticon. It not-so-simply provides a more generalized model then either of those two efforts.

But the basic idea is the same: capture a query against a given NSF and turn it into an agent process that can deliver more robust information than the HTTP renderer knows how to provide. That is a great thing for Domino to have.

There's probably only a handful of people that know how to leverage that capability -- just like there's only a handful of people that know how to leverage ?ReadViewEntries. But look what Jack Ratcliff and Rich Waters have accomplished with that!

So I hope we can take Chris's abstraction here and turn it into some directly useful facilities. Web-based DbLookups spring to mind as an easy-to-implement solution.

But more generally...

Well, I'll just say that the idea of building templates to wrap HTTP return data is not a waste of time in relation to the long-term plans I've heard.

Is that vague enough?

Heh... if you hear that I got kicked out of the DP programs, then the answer is: no, it wasn't vague enough. :-D

@Chris, if you've not already done so, try and get your hands on a copy of "RESTful Web Services" by Leonard Richardson & Sam Ruby.

http://safari.oreilly.com/9780596529260

I think you've got an interesting project here, but saying that it's a REST api is going to bring a lot of completely unnecessary baggage of expectation and confusion to the party. Just say it's a simplified http based api and you sidestep the whole can of worms.

Anywho, best of luck with this. :)