required Java version (Was: Questions about the release instructions)

[David Crossley <cr...@apache.org>]

Ferdinand Soethe wrote:
> David Crossley wrote:
> 
> >> >>- Set your Java version to be the lowest specified of our supported
> >> >>versions.
> >> >
> >> >Where is the definitive info on the supported versions? The FAQ says
> >> >we need Java 1.4 or better. Does that mean it has to be 1.4.0 or can
> >> >it be 1.4.x
> >> 
> >> 1.4.0
> 
> > Oooh, glad you asked that question. I was just going to use 1.4.1
> > Last time i did not use 1.3.0 rather 1.3.1
> 
> I think that we should test with 1.4 in its most current
> fix-version and update our requirements to reflect that.
> 
> If we test against 1.4.0 we might find bugs that result from bugs in
> 1.4.0 and have been fixed in the more recent versions. And we might
> not find bugs that have been introduced with these fixes.
> 
> So we are hurting people that do what you should be doing (update your
> java regularly to the latest fix version) to what end?

I have similar concerns. Using 1.4.0 sounds very risky.
However we could not expect everyone to use the most recent 1.4.x

So would a compromise of 1.4.1 be the most sensible?

--David

Re: required Java version (Was: Questions about the release instructions)

[Juan Jose Pablos <ch...@apache.org>]

Ross Gardler wrote:
> 
> I would prefer to see us test on 1.4.0 but since I'm not doing the 
> release process I'm happy with 1.4.1 if it makes your lives easier (it 
> actually makes mine harder as I don't have 1.4.1, I have 1.4.0 and 
> 1.4.2, but not 1.4.1).

I think that this is trivial. When I had been compiled cocoon for 
forrest I have been using the minimum major (1.4 for 2.2-dev) and the 
higher patched version (today would be: 1.4.2_08)

But I am happy to whatever the person who does the job choose.

If there is a company with millions of dollars that can not afford to 
upgrade their system from 1.4.0 to 1.4.2_08 then they have one choice:

As they have all the source: compile it themselves.

If that fails they can report a bug and we will follow this discussion 
then :-)

WDYT?

Re: required Java version (Was: Questions about the release instructions)

["Gregor J. Rothfuss" <gr...@apache.org>]

Ferdinand Soethe wrote:

> Backwards compatibility obviously cannot mean that a version with bugs
> (that might cause problems with Forrest) will not cause this problem
> because the fix version (where that bug is fixed) is supposed to be
> backward compatible.
> 
> It would only mean that this fixed version should also work with a
> Forrest that works with the buggy earlier version.
> 
> So, we would be creating a workaround for a Java bug that the fixed
> version might no require any longer. => do double work.

in years of java development, i never hit a bug in the jdk. pretty 
remote probability, imo.

Re: required Java version (Was: Questions about the release instructions)

[Ferdinand Soethe <sa...@soethe.net>]

To make that clear before I argue any further:

I accept that there are at least two people in favor of using 1.4.1
and so we should do that of course.

But I'd still like to argue my case :-)


Ross Gardler wrote:

>> 1. If we test against anything other than the most recent fix version
>>    of Java we might have to develop fixes for problems that have
>>    already been solved by the Java folks.

> Minor releases of Java are intended to be backward compatible, we have
> to go with the assumption that they are.

Backwards compatibility obviously cannot mean that a version with bugs
(that might cause problems with Forrest) will not cause this problem
because the fix version (where that bug is fixed) is supposed to be
backward compatible.

It would only mean that this fixed version should also work with a
Forrest that works with the buggy earlier version.

So, we would be creating a workaround for a Java bug that the fixed
version might no require any longer. => do double work.

> In addition, Forrest is developed and used in many different versions (I
> have clients in 1.4.1, 1.4.2 and 1.5.0 environment, I am sure many other
> devs use it in multiple environments too. In other words it is being
> tested all the time in different environments. The testing in the 
> release process is just a little more rugged.

>> 2. Installing the latest Java fix might solve
>>    problems with other Java software and close up some security holes, so
>>    people would benefit in more than one respect.

> You must understand that Forrest is used in production environments 
> where upgrading Java is not a trivial thing. If we insist on "the 
> latest" fix version we will be preventing people from upgrading to 
> Forrest 0.7.

From my working with a few huge bank organizations I observed that
they will not do regular upgrades when it comes to feature releases
such as 1.5 or 2.x.

I also found that it is very hard to predict which feature version
they will use at what time.

But they usually try to follow fix releases (third decimal) quite
closely because - as you said - these don't introduce new
features/bugs, cause few compatibility problems and usually fix
problems.

> In other words, Forrest is not that important to users that they will
> spend hundreds or even thousands of dollars/euros/pounds upgrading their
> environment for it.

>> 3. Using the latest fix is usually the common denominator for all
>>    software using common layers like Java.

> That is not my experience. If that was the case we would be using Jave
> 1.5,

Seconds decimals in my understanding are minor feature releases which
means new features, new problems.

> but the reality is that most companies are always at least pne 
> major release behind the curve because major releases usually introduce
> bugs.

Exactly!

> I even have a client still on Java 1.2.2 (not using Forrest though).

>> So short of knowing that the latest fix has such a big problem, I'd
>> suggest to test against that unless somebody comes up with more good
>> reasons why not to.

> I'm -1 on testing any later than 1.4.1, my resons are above. Like all
> veto's I'll remove it under community pressure and good argument, but
> none of those presented above are strong enough for me at this time. Soryy.

Fine by me.

--
Ferdinand Soethe

Re: required Java version (Was: Questions about the release instructions)

[Ross Gardler <rg...@apache.org>]

Ferdinand Soethe wrote:
> Sorry, I completely missed that response in the flurry of postings.
> 
> David Crossley wrote:
> 
> 
>>I have similar concerns. Using 1.4.0 sounds very risky.
> 
> 
>>However we could not expect everyone to use the most recent 1.4.x
>>So would a compromise of 1.4.1 be the most sensible?
> 
> 
> I disagree. For several reasons:

I disagree with your disagreement ;-)

I would prefer to see us test on 1.4.0 but since I'm not doing the 
release process I'm happy with 1.4.1 if it makes your lives easier (it 
actually makes mine harder as I don't have 1.4.1, I have 1.4.0 and 
1.4.2, but not 1.4.1).

> 1. If we test against anything other than the most recent fix version
>    of Java we might have to develop fixes for problems that have
>    already been solved by the Java folks.

Minor releases of Java are intended to be backward compatible, we have 
to go with the assumption that they are.

In addition, Forrest is developed and used in many different versions (I 
have clients in 1.4.1, 1.4.2 and 1.5.0 environment, I am sure many other 
devs use it in multiple environments too. In other words it is being 
tested all the time in different environments. The testing in the 
release process is just a little more rugged.

> 2. Installing the latest Java fix might solve
>    problems with other Java software and close up some security holes, so
>    people would benefit in more than one respect.

You must understand that Forrest is used in production environments 
where upgrading Java is not a trivial thing. If we insist on "the 
latest" fix version we will be preventing people from upgrading to 
Forrest 0.7.

In other words, Forrest is not that important to users that they will 
spend hundreds or even thousands of dollars/euros/pounds upgrading their 
environment for it.

> 3. Using the latest fix is usually the common denominator for all
>    software using common layers like Java.

That is not my experience. If that was the case we would be using Jave 
1.5, but the reality is that most companies are always at least pne 
major release behind the curve because major releases usually introduce 
bugs. I even have a client still on Java 1.2.2 (not using Forrest though).

> So short of knowing that the latest fix has such a big problem, I'd
> suggest to test against that unless somebody comes up with more good
> reasons why not to.

I'm -1 on testing any later than 1.4.1, my resons are above. Like all 
veto's I'll remove it under community pressure and good argument, but 
none of those presented above are strong enough for me at this time. Soryy.

Ross

Re: required Java version (Was: Questions about the release instructions)

[Ferdinand Soethe <sa...@soethe.net>]

Sorry, I completely missed that response in the flurry of postings.

David Crossley wrote:

> I have similar concerns. Using 1.4.0 sounds very risky.

> However we could not expect everyone to use the most recent 1.4.x
> So would a compromise of 1.4.1 be the most sensible?

I disagree. For several reasons:

1. If we test against anything other than the most recent fix version
   of Java we might have to develop fixes for problems that have
   already been solved by the Java folks.
   
   To me that sounds like a waste of our resources. And even though we
   don't get paid, I still value these resources higher than the time
   people have to spend on downloading and installing the latest fix
   (which they really only have to if and only _if_ there is indeed a
   rare problem).

2. Installing the latest Java fix might solve
   problems with other Java software and close up some security holes, so
   people would benefit in more than one respect.

3. Using the latest fix is usually the common denominator for all
   software using common layers like Java. Unless - like with Lotus Notes
   at some point - the latest version is known to have introduced some
   big problem. But if that is the case, you really have a mess
   anyway.

So short of knowing that the latest fix has such a big problem, I'd
suggest to test against that unless somebody comes up with more good
reasons why not to.

But if there is no majority for that, I'd always take the best that is
latest fix we can agree on.

--
Ferdinand Soethe