On Thu, Feb 14, 2013 at 2:59 PM, John Carlson <[email protected]> wrote:

> Miles wrote:
> > For the most part, if you want to retrieve JSON data, you need to use
> some protocol - the RESTful model is to address the data with a URL, and
> use an HTTP GET.  If you want to upload or replace some data, you use HTTP
> PUT, and to delete it you use HTTP DELETE. That's about as simple as you
> can get.
>
> What I am saying is simple is TOO SLOW, especially when dealing with
> individual ids.  Why are you rejecting me?
>
You are speculating that it is too slow. Have you benchmarked it? That's a
rhetorical question. Of course you haven't. You haven't even stated a
specific application for which this is allegedly "too slow". Even without
HTTP pipelining, I'd bet that operating on individual IDs is fast enough
for most applications. With HTTP pipelining, performing a hundred PUTs vs.
one POST with a hundred records might only be a small factor on bandwidth,
with a negligible effect on latency.

If you have some model for mutable views (bidirectional computation,
lenses) then you could also represent a view-function as the URL/ID and PUT
a single value to that view to update hundreds of individual records. This
is a bit esoteric, of course; pervasive support for bidirectional
computation isn't readily available outside of research systems (Boomerang,
RDP, etc.).
_______________________________________________
fonc mailing list
[email protected]
http://vpri.org/mailman/listinfo/fonc

Reply via email to