7:36 p.m.
Hi guys,
So I have been thinking about websockets, and I think we should make
some changes.
At the moment we essentially have 3 different API's, the low level XNIO
API, a Undertow native high level API, and a JSR-356 API (that will
hopefully stop changing constantly now).
I think 3 API's is a bit excessive, and we should just have our XNIO
Channel based low level API, and the JSR-356 API.
I am also thinking we should split it out of Undertow entirely, as
really the only undertow dependency is the handshake handler. This
should make it easier for us to ship a JSR-356 standalone web socket
client.
Thoughts?
Stuart
6:54 a.m.
The XNIO API allows you to send and receive messages using XNIO StreamSinkChannel and
StreamSourceChannel.
The high level API just provides an abstraction over that that is easier to use but less
flexible, such as allowing you to just send/receive text frames as Strings. The JSR-356
provides an API that performs basically the same function, with some extra stuff such as
annotated endpoints and encoders and decoders thrown in.
----- Original Message -----
From: "Heiko Braun" <hbraun(a)redhat.com>
To: "Stuart Douglas" <stuart.w.douglas(a)gmail.com>
Cc: undertow-dev(a)lists.jboss.org
Sent: Friday, 12 April, 2013 5:04:46 PM
Subject: Re: [undertow-dev] Websockets
What's the difference between the XNIO and the high level one?
On Apr 12, 2013, at 2:36 AM, Stuart Douglas < stuart.w.douglas(a)gmail.com >
wrote:
I think 3 API's is a bit excessive, and we should just have our XNIO
Channel based low level API, and the JSR-356 API.
_______________________________________________
undertow-dev mailing list
undertow-dev(a)lists.jboss.org
https://lists.jboss.org/mailman/listinfo/undertow-dev
10:49 a.m.
I have had a conversation via e-mail with Mark Little about this particular aspect of
Undertow. The reason that conversation started was because Mark asked me about the status
of a particular TAG issue, which was about Netty vs. XNIO.
I had thought that we had agreed, in regards to that TAG issue that XNIO and Netty served
two different purposes, and that if you needed something like websockets (already in
Netty), that we would use Netty, and if you didn't need additional functionality
outside of low level NIO functionality, you would use XNIO.
So, this is a little off topic, but I'm wondering how we ended up not using Netty for
the websockets for Undertow in the first place. It seems like we have built a lot of
duplicate functionality for what was already available in Netty.
So, perhaps this should branch off into a different thread, but we need to have this
conversation.
Andy
----- Original Message -----
From: "Stuart Douglas" <stuart.w.douglas(a)gmail.com>
To: undertow-dev(a)lists.jboss.org
Sent: Thursday, April 11, 2013 6:36:42 PM
Subject: [undertow-dev] Websockets
Hi guys,
So I have been thinking about websockets, and I think we should make
some changes.
At the moment we essentially have 3 different API's, the low level
XNIO
API, a Undertow native high level API, and a JSR-356 API (that will
hopefully stop changing constantly now).
I think 3 API's is a bit excessive, and we should just have our XNIO
Channel based low level API, and the JSR-356 API.
I am also thinking we should split it out of Undertow entirely, as
really the only undertow dependency is the handshake handler. This
should make it easier for us to ship a JSR-356 standalone web socket
client.
Thoughts?
Stuart
_______________________________________________
undertow-dev mailing list
undertow-dev(a)lists.jboss.org
https://lists.jboss.org/mailman/listinfo/undertow-dev
4651
days inactive
4654
days old
3 comments
4 participants
participants (4)
-
Andrig Miller -
Heiko Braun -
Stuart Douglas -
Stuart Douglas