Open vs Closed java on linux (was: Re: [Nottingham] What distro?)

Martin Garton martin at stupids.org
Thu May 25 10:23:44 BST 2006


On Thu, 2006-05-25 at 09:59 +0100, Michael Erskine wrote:
> On Thursday 25 May 2006 09:26, Martin Garton wrote:
> > I haven't tried netbeans, but last time I tried eclipse (on FC4 I think)
> > it ran slightly slower than under sun java, but started up faster.  It
> > also appears to be highly dependent on gcj version and gnu classpath
> > version.  It might be worth trying again.
> 
> You ain't joking! Eclipse under gcj is grossly slow and just plain does not 
> work for debugging. And the answer in the mailing lists? Just use the Sun 
> JRE! 

Okay, point taken - I hadn't tried debugging. It was perfectly usable
for the couple of days I used it though, editing, compiling and running
code at an acceptable speed.

> > Progress of open java is interesting.  One reason that open java has
> > taken (and still is taking) longer to progress on linux than open .NET
> > is that there is closed java available and people choose that by default
> > because it's more convenient, so the open implementations don't get much
> > testing.  If everyone chose an open implementation (as they are forced
> > to do for .NET on linux) and took a few minutes to report bugs if it
> > didn't work for them, we would have a better faster more robust open
> > java in no time.
> 
> Like a lot of people, when lucky enough to be presented with a choice, and 
> when I just want to get on with the job, I tend to go for whatever works 
> first or is least broken.

Exactly. That's what I meant by "people choose that by default because
it's more convenient"   If convenience is your aim, it's definitely the
right choice.

> Whilst it's true that bug reporting might result in bugs being fixed, I expect 
> that a vague bug report of "gcj is too slow to run eclipse" will just be 
> ignored or be followed up with posts for people with more time: "do this", 
> "try this", "we need this obscure info", "use the CVS version, everybody else 
> does", "just rebuild your kernel with these obscure flags", "delete this part 
> of your install and see if it's any better" -- really, this is what you'll 
> get.

A admit that for performance bug reporting you would need some basic
profiling skills to really give any useful info.  There are plenty of
other bugs in open java though (!) which anyone could submit reports
about and which I expect would really help.   In the early mono days I
submitted various bug reports and was amazed at the response.

>  And while there are people out there who have the time and expertise to 
> do all these things, there are many others who just want to get on with the 
> job.

Agreed.  I'm not trying to say "closed java is bad and evil - don't use
it",  I'm saying "if you care about open java and want it to be better,
try it occasionally with your favorite java apps and report bugs if you
can."

-- 
Martin.





More information about the Nottingham mailing list