Hi,
on topic:
> I'm not sure this should be fixed by adding support for mulitple
> interceptors (on a ProcessorTask, e.g.). Most of the changes so far
> involve altering deprecated classes which is a little
> contradictory. In 2.x, the planned approach would be to use the
> port-unification/ProtocolFilter approach and I think we should do
> that for 1.9.x as well. It's a much, much clean approach and could
> be done with relatively little change to existing code. It'd just
> be another PF in the chain that would halt any further execution of
> the chain. So it'd be the same basic logic as the current suggested
> approach without all the ugly hackery going on to retrofit this
> update in a non-breaking fashion. As it is now, it's looking more
> and more like a breaking API change on the 1.9.x tree and that's
> something we're striving to avoid. Thoughts?
You're right, this issue is basically addressed just to 1.9.x branch.
The extension will help us to support multiple protocols/frameworks on
top of HTTP. Now, Grizzly HTTP server is an atomic block, so there is
no other way to register some custom HTTP protocol implementation
inside it than implementing own HTTP Interceptor.
In 2.0 (may be not first release) we will split HTTP/Web container
implementation into separate Filters: HTTPFilter and Web container
Filter. HTTP Filter will be just responsible for parsing HTTP request
so developers will be able to insert any custom protocol Filter next
to HTTP Filter in order to implement custom HTTP based protocol.
off topic:
I will just write my thoughts and won't make any other comment on that.
> I will never work with 2.0. I disagree with its design.
It's your right. So far I heard no concrete proposal or comment from
you on 2.0 design.
> Its of interest that JFA dont like it either, wonder why he left
> anyhow ?
That's so absurd and not truthful comment, that I have nothing to say.
> He stated two times to me last year that 2.0 is "slow". last time
> was in december.
2.0 is still being developed and changed. We never tested 2.0 perf.
officially. So we can state it's slow or it's fast... only tests will
show that.
> Its ok that my hack for websocket integration is not viable, its sun
> management that ask for websocket, not me. im not intereted now.
> I can work for free at other projects that might have a more
> economical future.
> Its too unclear what oracle will do for me to spend any further time
> atm.
Good luck.
As I've mentioned before it's my first and last comment on off topic
part.
Thanks.
WBR,
Alexey.
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: dev-unsubscribe_at_grizzly.dev.java.net
> For additional commands, e-mail: dev-help_at_grizzly.dev.java.net
>