Fwd: O'Reilly invites you to Mashed 2008

[Paul Makepeace]

On Thu, Jun 19, 2008 at 1:23 PM, Nicholas Clark <nick at ccl4.org> wrote:
> On Thu, Jun 19, 2008 at 12:20:01PM +0100, Paul Makepeace wrote:
>> On Thu, Jun 19, 2008 at 11:45 AM, Nicholas Clark <nick at ccl4.org> wrote:
>> > On Thu, Jun 19, 2008 at 10:31:07AM +0100, peter at dragonstaff.com wrote:
>> >
>> >> What do you make of the following?
>> >>
>> >> http://www.slideshare.net/bquinn/xtech-2008-bbc-tech-refresh-and-identity-brendan-quinn-and-ben-smith
>> >>
>> >> http://phpimpact.wordpress.com/2008/06/02/bbc-new-infrastructure-java-and-php/
>> >
>> > I keep wondering how much more hardware it's going to take to serve pages
>> > that way,
>>
>> Presumably fewer since Java is much less resource hungry than Perl.
>
> Is that general wisdom, or direct personal experience? And has anyone actually
> quantified it?

Plenty of people have quantified it. JIT-compiled (raw) Java is
_fast_. It's close to C++; I've even seen some bizarre examples where
it's come out quicker than C owing to compiler being cleverer with the
available information (pointers being hard to optimize & generalize
about).

I think anecdotally Java can consume quite a lot of memory but I'm
less familiar with this side. (Then again, mod_perl isn't lite on RAM
either...)

> For a straw poll of 1 instance, I remember much the opposite - once Fotango
> started using Java for something, the dev server went to hell.  Previously
> it had been happy to support all the developers running stuff using
> mod-perl, now it was being brought to its knees by the memory demands of the
> servers being run by a small subset of developers. However, to be fair, I'm
> not sure that the Java servers were ever well configured for their
> development role, and may well have been spawning too many threads too early.

Yeah, there are some Java servers that are desperately bad, the J2EE
stuff is/was like a dog. I've also heard of horror stories moving to
Java appservers. (Though not with Tomcat, which is open source, go
figure...)

> [And is the real answer - it depends how well the code is written, and how
> well it plays to the strengths of the language?]
>
>> Choice of PHP seems a bit puzzling to me; curious what the rationale
>> there was, given their using Java. They're not using a lot of the
>> "good" bit of Zend either.
>
> The "good" bit being what? A damn fast server for running PHP bytecode?

I was referring to the other middleware type stuff in Zend. Having a
fast server is good though, and I suspect that's why they chose it.

P

>> > and how long before the system actually goes into production use.
>>
>> Bloody ages, if Siemens has anything to do with it. Anyone know what
>> the story is there?
>
> No idea. But it keeps making me thing of http://despair.com/consulting.html
>
> Nicholas Clark
>