Background: We have a quite a few customers who use really unreliable
internet connections [e.g. 3g connections in bangladesh]. Often times
these requests can take a long time to complete if they are in a
country with really poor internet, but the users keep on doing things
and then requests pile up. Obviously browsers only let two go at a
time, and then we start to run into timeout issues.

Recently I implemented a system that patched proxycreator to allow me
to control request error handling, optionally resending the serialized
request when requests fail, trapping exceptions, logging, performance
etc. This is needed in production because of weird issues with IE &
firefox that happen on some customers setups, and works flawlessly.

I was thinking that there would be an opportunity for a global request
queue, which would queue requests (have only two in transit at any one
time), but would allow us to pool requests (e.g. send multiple
payloads at once to the server if there are a lot of requests in the
queue). A simple servlet filter would be able to process the combined
request and return the results to a special pooled callback, which
then calls the relevant client callbacks.

Given that on really bad internet connections latency is much worse
than throughput I think this would cause a huge performance
improvement to our users. I have lots's of data from request logs from
the client side showing that this is a problem, and doubt I am alone
in this.

My question is:
1) Would people be interested in my proxycreator patch, for potential
inclusion into gwt core?
2) Does request pooling sound like it's something that would be of use
in general? Is this something the gwt team would be interested in
working with?

-- 
http://groups.google.com/group/Google-Web-Toolkit-Contributors

Reply via email to