async multi requests in the Java client

[Camille Fournier <ca...@apache.org>]

Can anyone jog my memory about why we didn't implement multi requests with
async callbacks for the Java client like we did in the C client?
I think it's not a great precedent to set because now we have bugs that
have been discovered in the implementation of multi on the server side that
are easily reproducible only via the C client, see:
https://issues.apache.org/jira/browse/ZOOKEEPER-1624


Multi was a big project so I can understand why we let this go for the
first impl, but unless there's a technical reason not to support it in the
Java client it would be a good fix for us to put it in there as well,
especially in light of this issue.

Thoughts?

C

Re: async multi requests in the Java client

[Ted Dunning <te...@gmail.com>]

Yeah.  It was an oversight.  I did the java client and I ten not to think much about the asynchronous interfaces so I never designed anything along those lines.  

Should be hard to add.  

Sent from my iPhone

On Feb 1, 2013, at 2:04 PM, Marshall McMullen <ma...@gmail.com> wrote:

> As far as I know it was not intentional.
> 
> My best guess here is that is was due to different people working on
> different parts of Multi. Ted did the original Java client work but didn't
> have the time to do the C client work and my company really needed that
> functionality so I volunteered to implement it. We never discussed the
> async interface. I only implemented it because the C client necessitates
> it. All of the synchronous interfaces are just wrappers around the
> asynchronous ones.
> 
> I think this was definitely an oversight. At a minimum it seems during the
> many code reviews the Multi code got this should have been noticed..
> 
> 
> On Fri, Feb 1, 2013 at 2:45 PM, Camille Fournier <ca...@apache.org> wrote:
> 
>> Thawan's followup on that ticket pointed me to this open issue:
>> https://issues.apache.org/jira/browse/ZOOKEEPER-1572
>> 
>> So that resolves it, which is great.
>> My general call-out on divergent functionality away from the Java client
>> stands for future reference but looks like we're good for this one.
>> 
>> C
>> 
>> 
>> On Fri, Feb 1, 2013 at 4:42 PM, Camille Fournier <ca...@apache.org>
>> wrote:
>> 
>>> Can anyone jog my memory about why we didn't implement multi requests
>> with
>>> async callbacks for the Java client like we did in the C client?
>>> I think it's not a great precedent to set because now we have bugs that
>>> have been discovered in the implementation of multi on the server side
>> that
>>> are easily reproducible only via the C client, see:
>>> https://issues.apache.org/jira/browse/ZOOKEEPER-1624
>>> 
>>> 
>>> Multi was a big project so I can understand why we let this go for the
>>> first impl, but unless there's a technical reason not to support it in
>> the
>>> Java client it would be a good fix for us to put it in there as well,
>>> especially in light of this issue.
>>> 
>>> Thoughts?
>>> 
>>> C
>>> 
>>> 
>>> 
>> 

Re: async multi requests in the Java client

[Marshall McMullen <ma...@gmail.com>]

As far as I know it was not intentional.

My best guess here is that is was due to different people working on
different parts of Multi. Ted did the original Java client work but didn't
have the time to do the C client work and my company really needed that
functionality so I volunteered to implement it. We never discussed the
async interface. I only implemented it because the C client necessitates
it. All of the synchronous interfaces are just wrappers around the
asynchronous ones.

I think this was definitely an oversight. At a minimum it seems during the
many code reviews the Multi code got this should have been noticed..


On Fri, Feb 1, 2013 at 2:45 PM, Camille Fournier <ca...@apache.org> wrote:

> Thawan's followup on that ticket pointed me to this open issue:
> https://issues.apache.org/jira/browse/ZOOKEEPER-1572
>
> So that resolves it, which is great.
> My general call-out on divergent functionality away from the Java client
> stands for future reference but looks like we're good for this one.
>
> C
>
>
> On Fri, Feb 1, 2013 at 4:42 PM, Camille Fournier <ca...@apache.org>
> wrote:
>
> > Can anyone jog my memory about why we didn't implement multi requests
> with
> > async callbacks for the Java client like we did in the C client?
> > I think it's not a great precedent to set because now we have bugs that
> > have been discovered in the implementation of multi on the server side
> that
> > are easily reproducible only via the C client, see:
> > https://issues.apache.org/jira/browse/ZOOKEEPER-1624
> >
> >
> > Multi was a big project so I can understand why we let this go for the
> > first impl, but unless there's a technical reason not to support it in
> the
> > Java client it would be a good fix for us to put it in there as well,
> > especially in light of this issue.
> >
> > Thoughts?
> >
> > C
> >
> >
> >
>

Re: async multi requests in the Java client

[Camille Fournier <ca...@apache.org>]

Thawan's followup on that ticket pointed me to this open issue:
https://issues.apache.org/jira/browse/ZOOKEEPER-1572

So that resolves it, which is great.
My general call-out on divergent functionality away from the Java client
stands for future reference but looks like we're good for this one.

C


On Fri, Feb 1, 2013 at 4:42 PM, Camille Fournier <ca...@apache.org> wrote:

> Can anyone jog my memory about why we didn't implement multi requests with
> async callbacks for the Java client like we did in the C client?
> I think it's not a great precedent to set because now we have bugs that
> have been discovered in the implementation of multi on the server side that
> are easily reproducible only via the C client, see:
> https://issues.apache.org/jira/browse/ZOOKEEPER-1624
>
>
> Multi was a big project so I can understand why we let this go for the
> first impl, but unless there's a technical reason not to support it in the
> Java client it would be a good fix for us to put it in there as well,
> especially in light of this issue.
>
> Thoughts?
>
> C
>
>
>