commons-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Alex Karasulu" <>
Subject RE: Commons server infrastructure proposal
Date Mon, 12 Apr 2004 17:16:35 GMT

> From: Oleg Kalnichevski []
> With so many REALLY controversial changes to Java language long may be
> our wait for 1.5 to be widely adopted. Java 1.5 requirement may be quite
> a hindrance and may play against the project in the initial phase

Oh I got that feel too.  

> > mechanism we need.  If forced to we should be able to implement a
> > SSLSocketChannel for 1.4 no?
> >

> I entertained this idea for while and finally decided against it. I
> sense standard Channel classes should be using platform dependent calls
> though JNI to be efficient. One can certainly simulate 'select' using

We 'should' obviously stay away from JNI but it can be a temporary 
option.  NIO does allow one to plugin their own implementations.

> good ol' sleeping for a while/polling technique (this is what I ended up
> doing). The question is why one would need NIO in the first place.

Non-blocking IO is the primary reason for wrestling with NIO.

> brought out a strong conviction that for Java 1.4 NIO with all its
> stability problems does bring too much to the table. I hope NIO will
> mature by 1.5 release, however, I can hardly imagine myself using 1.5 in
> the foreseeable future.

Many will agree with you there.  The NIO selector mechanism is non-
trivial and requires some time getting used to and 1.5 is certain 
something that will require people to test the waters on.

I'm patient but would like to have these discussions and some 
trajectory towards building this framework Mladen is referring to.  
We can use many design tactics to cope with these issues we're 
currently discussing.  I'm confident it can be done right when we 
encounter them one by one.

The questions with my ideas on em are:

1). What do we need to do to kick it off?  
	Let's start playing in the commons sandbox if that's cool.

2). There is a lot of overlap so should it start from scratch?
	Yeah why not we can draw from the experience and code of other 
      Apache projects as we see fit but I don't mind starting clean.

3). All the interested parties seem to Apache members or committers so do we
need to go through incubation?
     Don't think so but I'll let the wiser ASF folks answer this one.



To unsubscribe, e-mail:
For additional commands, e-mail:

View raw message