[OSM-dev] PUT /api/0.6/{node|way|relation}/{id|create}

SteveC steve at asklater.com
Tue Jun 10 18:52:45 BST 2008


On 10 Jun 2008, at 18:46, Stefan de Konink wrote:

> SteveC schreef:
>> On 9 Jun 2008, at 20:26, Stefan de Konink wrote:
>>> The payload should be validated anyhow. And going to the 'first'  
>>> child
>>> is an operation that should happen too.
>>>
>>> So if this is only because 'rails needs it' I hope the C++ guys or  
>>> my
>>> code can replace this ugly duplication soon.
>> really, there are more important things to 'fix'.
>>    http://en.wikipedia.org/wiki/REST
>
> I thought you were the non-academic guy.

Well, I'm certifiably stupid.

>> it makes the routing work
>> it makes each controller separable
>> it looks nice
>> it has some logic behind it
>
> It is not required, thus the high performance API server could  
> ignore it ;) The routing itself could also be done based on the  
> request type. GET/POST/PUT.

look this project is not about the best most awesomely fast technology  
on the planet, it's about simple things that work. What you're doing  
it making it more complicated for no particular reason, other than to  
save what, 3-8 bytes in a HTTP request?

>> all you're advocating is like going back to something awful like  
>> XMLRPC with a single endpoint.
>
> Am I? I'm just looking at the requirement and reducing the amount of  
> checks, back to what is actually required for 'creates/updates' but  
> keep it as pretty outside as you want, I guess mod_rewrite got a  
> usable boost by Wikipedia too...

That's all wonderful, but balance your gains with your costs. Yes you  
get a server that is three microseconds faster, but you loose a simple  
clean and consistent API.

Best

Steve





More information about the dev mailing list