# Lumping vs. Splitting (REST)

**URL:** <https://www.ruby-forum.com/t/lumping-vs-splitting-rest/80266>\
**Category:** Rails\
**Created:** [December 22, 2006, 10:56pm UTC](https://www.ruby-forum.com/t/lumping-vs-splitting-rest/80266 "2006-12-22T22:56:59Z")\
**Posts on this page:** 4\
**Page:** 1

<div class="post-metadata">

**Author:** ![thehobbster](https://www.ruby-forum.com/letter_avatar_proxy/v4/letter/t/8e8cbc/32.png) [@thehobbster](https://www.ruby-forum.com/u/thehobbster)\
**Post date:** [December 22, 2006, 10:56pm UTC](https://www.ruby-forum.com/t/lumping-vs-splitting-rest/80266/1 "2006-12-22T22:56:59Z")

</div>

Hi Everyone,

Working on a REST-based web application, and I need some ammo in  
dealing with my supervisors and contractors.

Is it better to:

Method 1:  
Take in one large XML-based POST request, which contains 1st, 2nd  
and 3rd level elements, and build the associated child and grandchild  
objects in the parent controller,

Method 2:  
Instead use a number of smaller XML-based POST requests targeted at  
the appropriate child resources?

In this example, we have Routes, Stops and Visitors.

Using Method 1, the POST request would be aimed at the Routes  
controller, and the routes controller would parse the XML and build  
the stops and the visitors records. Using Method 2, each prior node  
of the XML document would make up a separate request, so there would  
be POST request against the routes controller, then for each route a  
series of POST requests against the Stops controller, and finally for  
each stop, a series of POST requests against the visitors controller.

From a development standpoint, I’m inclined to go for Method 2, but  
I’m facing some resistance. Can anyone offer some decent arguments  
in support in favor of Method 2? Or, if I’m wrong and Method 1 is  
the way to go, offer decent arguments in favor of Method 1 so that I  
understand where I’m confused?

Thanks!

-Jared

---

<div class="post-metadata">

**Author:** ![thehobbster](https://www.ruby-forum.com/letter_avatar_proxy/v4/letter/t/8e8cbc/32.png) [@thehobbster](https://www.ruby-forum.com/u/thehobbster)\
**Post date:** [December 22, 2006, 11:07pm UTC](https://www.ruby-forum.com/t/lumping-vs-splitting-rest/80266/2 "2006-12-22T23:07:02Z")

</div>

If you are thinking of implementing UD for the second level models,  
then method 2 definitely makes sense (ie, since you’re coding UD,  
then C would likely belong there as well). R might be reserved for  
the high level if it is the primary point of contact.

Hopefully of some help.

Jodi

---

<div class="post-metadata">

**Author:** ![thehobbster](https://www.ruby-forum.com/letter_avatar_proxy/v4/letter/t/8e8cbc/32.png) [@thehobbster](https://www.ruby-forum.com/u/thehobbster)\
**Post date:** [December 22, 2006, 11:21pm UTC](https://www.ruby-forum.com/t/lumping-vs-splitting-rest/80266/3 "2006-12-22T23:21:03Z")

</div>

On 12/22/06, Jared H. [removed\_email\_address@domain.invalid](mailto:removed_email_address@domain.invalid) wrote:

> be POST request against the routes controller, then for each route a  
> series of POST requests against the Stops controller, and finally for  
> each stop, a series of POST requests against the visitors controller.
> 
> From a development standpoint, I’m inclined to go for Method 2, but  
> I’m facing some resistance. Can anyone offer some decent arguments  
> in support in favor of Method 2? Or, if I’m wrong and Method 1 is  
> the way to go, offer decent arguments in favor of Method 1 so that I  
> understand where I’m confused?

Hi Jared. I find it easier to build and test such an API using method  
2.

Once you have things working and well-tested, you can create aggregate  
resources in the next iteration of your project.

If you’re wrapping the API with a client library like Active Resource,  
the  
difference in REST style will be transparent to the user anyway.

jeremy

---

<div class="post-metadata">

**Author:** ![thehobbster](https://www.ruby-forum.com/letter_avatar_proxy/v4/letter/t/8e8cbc/32.png) [@thehobbster](https://www.ruby-forum.com/u/thehobbster)\
**Post date:** [December 23, 2006, 9:59pm UTC](https://www.ruby-forum.com/t/lumping-vs-splitting-rest/80266/4 "2006-12-23T21:59:45Z")

</div>

i like method 2, i would design things from that perspective, once all  
is tested and functional build method 1 as a convenience method that  
parses the entire xml doc and delegates each task to the appropriate  
part of method 2 you have already built.

if the supervisors/contractors are non-technical, they are probably  
thinking that one request doing the entire task is easier for whomever  
is going to consume this service. Explain to them the benefits from the  
end users perspective of a design around method 2.