I'll bite I guess
RPC-Auth:
No specific comments, but the design looks sound, and should be easy
enough to tie in to authentication schemes on the server side.
+1
oh, in your simple login scenario, you have svc.login("foo", "foo",
someCallback); then your variable magically changes in the next line
to svn.restrictedMethod(someCallback); (but I'm sure you just did that
to see if anyone was actually reading)
RPC-Swizzle:
This looks like a reasonable solution to the issues currently seen
with collections and lists being sent back from Hibernate and such.
+1
de-RPC:
This one actually hits home as I've just been looking at ways to
improve some existing GWT code. I was performance testing the de-
serialization of JSON and RPC payloads for an Identical object graph.
The object graph in JSON format was 180K, and in RPC format was 68K.
It turned out that the de-serialization of the larger JSON data via
eval took about 20ms while the smaller (byte wise) RPC format took
almost half again as long 35ms. This appears to show that direct
evaluation is quite a bit faster, and that RPC could gain some
performance through its implementation. Hosted mode showed a
ridiculous speed advantage for direct evaluation 22ms vs 1480ms!
With the advent of overlay types for the javascript object, RPC is
losing a bit of its advantage, so adding a little performance boost,
as well as the possibility to use it on non-java back-ends could only
help.
-jason
On Sep 4, 2008, at 3:29 PM, BobV wrote:
>
> Is anyone up for a review?
>
> --
> Bob Vawter
> Google Web Toolkit Team
>
> >
--~--~---------~--~----~------------~-------~--~----~
http://groups.google.com/group/Google-Web-Toolkit-Contributors
-~----------~----~----~----~------~----~------~--~---
