Yeap, you’re pointing out exactly how high the tightrope is that we’re 
balancing on. :) The fact is, there are people out there right now- you might 
be one of them!- developing ZF components (components that play nicely with and 
likely depend on ZF ‘core’ components) that they would like to share with 
others either in their own company/projects and/or publically- and we haven’t 
addressed such components in any significant way as a framework. In fact, our 
proposal process right now is structured entirely around components on the 
‘core’ track, and many perfectly useful components will never be serious core 
candidates because they aren’t useful to most developers for any number of 
reasons. I know that people really appreciate the simple packaging that we 
currently have with ZF, and we’re not about to take away one of the biggest 
value propositions to our developers. On the other hand, we’re growing rapidly 
as a community and a software package, and what made sense yesterday isn’t 
necessarily practicable today. I doubt any of us want a full packaging system, 
so it will be a matter of finding the right compromise AFAIC tell. We haven’t 
had much time to talk about it lately since many of us are so focused on 1.5, 
but soon after the release we will pick the discussion up again on this list.

BTW, I greatly respect and sometimes even consult with my predecessors, but the 
course of ZF is set by the opinions of the ZF community *right now*. It may be 
hard to keep up sometimes, but that is pretty much the story of the software 
industry and OS community at large. And ZF has to keep up in its own right. ;)

 

,Wil

 

From: Kevin McArthur [mailto:[EMAIL PROTECTED] 
Sent: Tuesday, February 05, 2008 11:28 AM
To: Wil Sinclair
Cc: Bradley Holt; Elliot Anderson; Simone Carletti; [email protected]
Subject: Re: [fw-general] ZF Packaging

 

Wil,

I'll play devils advocate here.

First, before everything else one must consider deployment, packaging and 
installed base. One of the great failures with PDO, and possibly one of its 
successes, was with its separate driver architecture. Practically, splitting it 
makes sense, pragmatically however, it has lead to serious configuration 
headaches for a lot of users.

Zend Framework is the same, and your predecessors have expressed the importance 
in making sure that when a hosting package is selected that says 'Zend 
Framework v 1.X' it means all of it -- not a mix of components that will have 
to be detected at runtime. 

Worse, we could end up with independently versioned components aside from the 
core. Applications dependent on framework will then have to made even more 
intelligent, and deployment complexity increases significantly.

I would point you to review the archives here, as versioning and optional 
components have been discussed at length before.

There is something very powerful about Zend's currently unified deployment 
architecture.

Kevin






Wil Sinclair wrote: 

I generally like mootools’ and other JS libraries simplistic dependency model 
(AFAIC tell, the core lib is the only dependency allowed, so a general-purpose 
packaging system with tracking across full dependency graphs is not necessary). 
If we were to start distributing parts of ZF in a piecemeal fashion, I think 
that we would also see great benefit from a few basic rules aimed at 
drastically simplifying dependency management. While there is no immediate and 
significant runtime advantage that I can see here, we are interested in- and 
have been discussing- distributing at least one ‘lean and mean’ archive. There 
are several reasons for this, including lower load on our servers and- taking 
off my perfectly logical developer hat and putting on a more realistic 
marketing hat ;)- the fact that many developers and reviewers consider 
distribution size to be an important dimension on which to judge a framework. I 
don’t necessarily think that distribution size is a good indication of anything 
for a server-side framework beyond what you *can’t* expect to be included, such 
as sizable locale files which are very useful to our many international users 
but that add a MB or two to the current distribution of ZF (Thomas has done an 
excellent job getting these as small as possible while maintaining everything 
that makes them so useful in the first place), but I do think that the 
‘download only what you need’ distribution mechanism is both technically and 
philosophically compatible with ZF in its current state. We’ll probably be 
talking about this more once 1.5 is out the door.

 

,Wil

 

From: Bradley Holt [mailto:[EMAIL PROTECTED] 
Sent: Tuesday, February 05, 2008 7:04 AM
To: Elliot Anderson
Cc: Simone Carletti; [email protected]
Subject: Re: [fw-general] ZF Packaging

 

Elliot,

The main reason that mootools does this, in my understanding, is so that it can 
give you one JavaScript file to be included in your web page with only the 
components you need. There are performance advantages to this since you are 
only requiring the user's browser to get the JavaScript components it will 
need, not all of mootols. With Zend Framework, there is no performance 
advantage to only installing a handful of components. My understanding is that 
the performance hit comes when you require or include the component, not from 
it simply sitting on your web server. In other words, the main advantage of the 
pick-what-you-want download system doesn't apply when it comes to Zend 
Framework. The only advantage I can see is storage space, but have there been 
any complaints about that with Zend Framework?

Thanks,
Bradley

On Feb 5, 2008 6:26 AM, Elliot Anderson <[EMAIL PROTECTED]> wrote:

I'm a fan of the pick-what-you-want download system that Moo Tools has.

http://mootools.net/download







On Jan 29, 2008 6:04 AM, Simone Carletti <[EMAIL PROTECTED]> wrote:

On Jan 28, 2008 3:57 PM, Richard Thomas <[EMAIL PROTECTED]> wrote:

        zfdev.com is a community supported project that never really took off,
        It was never an "official" repository though.

 

Sorry Richard,
my misunderstanding. :)

Thanks for pointing it out.

Simone

 




-- 
Bradley Holt
[EMAIL PROTECTED]

Reply via email to