tapestry-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Alex Kotchnev <akoch...@gmail.com>
Subject Re: A plan for 5.3 / 5.4
Date Wed, 09 Feb 2011 14:18:06 GMT
I second Kristian's opinion - it seems that Howard's roadmap includes little
to no user facing features. In many respects it seems that there are a bunch
of user-facing rough spots like the Javascript framework, JPA, HTML5,
portlets that generate questions on the mailing list almost daily. I'm not
sure how many new users are sold on getting stared w/ the framework because
"it has yet more powerful metaprogramming features".

One area that I miss from the Grails world (that Play also seems to adopt)
is moving the "convention-over-configuration" from deep down in the code
into more visible parts of the system (e.g. into the filesystem). For
example, if you want to tweak the datasource configs, you just edit
"conf/DataSource.groovy", if you want to tweak the spring config, you tweak
"conf/resources.groovy" (or conf/resources.xml). If you want to add things
to the startup of the app, you tweak "Bootstrap.groovy". Much of that
simplicity is lost in T5 when I have to edit the already pretty long
AppModule.java in my app when I have to "contribute" into yet another
configuration point and implement yet another T5 interface. T5 already does
some of this w/ the special "components" and "pages" packages (and the
"entity" package w/ Hibernate support)

Another area that I also find massively useful in Grails is the built in
configurability. You get three environments (dev,test,prod in Config.groovy)
out of the box (and you can easily add more if needed) and in each
environment you can declare and override configuration points at will. Sure,
I've done something similar and very basic w/ Maven profiles, but it feels
like a bolt-on instead of being a nicely integrated part of the framework.

Testing T5 apps is still a sore spot. Sure, you can use EasyMock to mock out
a bunch of stuff and unit test components but you end up having to learn
about much more than you signed up for - having to mock out some T5 services
is not fun at all. Having better and more explicit support for various unit
testing tasks (e.g. once again, Grails has some excellent mocking utilities
that allow you to mock out various parts of the framework in unit tests that
are commonly needed - be it domain objects, controllers, services).  Testify
is an excellent addition but it's not a part of the framework and as such
can lag behind official T5 releases.

The items above are just the core-framework related things. From a
ecosystem/extension perspective the list goes on and on - support for schema
generation for hibernate, a generic data migration facility (e.g. based on
liquibase or yaml), standard security package etc, etc. Sure, I could
accomplish all of these on my own but the fact that the framework doesn't
cater to these common needs certainly doesn't help w/ picking it in the
first place. On one hand, I understand that T5 isn't a 'full stack' solution
("it's just a web framework for god's sake" - i hear something screaming in
my head) but when it tries to compete against others (such as Grails or
Play) it kinda has to be to stay competitive.

On the Gradle switchover - as far as I understand, I guess it doesn't have
to affect users of the framework , as long as the artifacts still continue
being exported into maven repositories (e.g. which I think Gradle does
well). A user of the framework can still continue using a maven archetype to
get started w/ Tapestry, it seems like there would be just one more way to
do so (w/ the equivalent of a Gradle project archetype). At the same time, I
don't see too many benefits of the switchover (but then I'm not a core
developer either, so my opinion doesn't hold much water) beyond having to
use a different tool for the build and having yet another build tool to
learn.

I appreciate Howard's drive to make the underpinnings of the framework yet
better , cleaner, and more powerful (e.g. the plastic work) ; however, at
least from the perspective of a "regular" user they don't have much
immediate value.  It seems to me that the framework periphery (as discussed
above) would be more valuable from a user's point of view. The core itself
has already been moving quite quickly (e.g. from 5.0->5.1->5.2 a whole bunch
of changes to the guts). Although these changes don't affect the clearly
delineated public APIs, they do affect many of the internals, which module
authors have no choice but do depend on, resulting into upgrade issues (e.g.
many people depend on third party modules like tynamo , testify
or chenillekit ). Maybe a little slowdown in core framework advances would
be nice for module authors as well.

Regards,

Alex K


On Wed, Feb 9, 2011 at 7:34 AM, Kristian Marinkovic <
kristian.marinkovic@porscheinformatik.at> wrote:

> i guess your roadmaps (igor, howard) are very promising.
>
> what i am missing is a decent (official) portlet support in T5. This was
> the main reason some of my customers refused to switch to T5 and chose JSF
> instead (with Liferay). They were all quite impressed by T5 features but
> they didn't want to spend their effort in maintaining a custom portlet
> solution. Those costumers are big companies that already have some kind of
> portlet support and want to reuse their infrastructure. Almost everytime
> we were presenting T5 and its advantages to some managers, someone had to
> ask: does it support portlets (event if they don't use it!). from my
> experience this is a big selling point.
>
> a new JS infrastructure / abstraction layer is overdue as this is the
> reason why there are no (at least not many) fancy JS enabled controls and
> widgets for T5. because everytime you think of creating one, you remember
> that T5 depends on prototype and you dont want to use it anymore. but
> writing T5 components that depend on JQuery, dojo, ... doesn't feel right
> too. i think it would be a big advantage if you could write js library
> agnostic components (with some additional adapters if necessary) and adapt
> to your customers needs.
>
> i'm still not convinced moving from gradle to maven is a good idea. for me
> maven works great and almost any it-company nowadays is familiar with it.
> we are building a custom OSGIfied version of T5.2 using maven. i don't
> know how to achieve the same result with gradle. Also from my point of
> view having a default project layout for all java projects (especially
> open source) is a good (not to say great) idea!
>
> g,
> kris
>
>
>
> Von:    Howard Lewis Ship <hlship@gmail.com>
> An:     Tapestry development <dev@tapestry.apache.org>
> Datum:  08.02.2011 22:34
> Betreff:        Re: A plan for 5.3 / 5.4
>
>
>
> Javassist vs. Plastic vs. ComponentClassTransformWorker
>
> Imagine if you could contribute a "classic"
> ComponentClassTransformWorker or some new interface for working with a
> Plastic class, both to the same service:
> ComponentClassTransformWorker.  The "classic" style would be
> transformed, via TypeCoercer, to the new form.  And, of course, we'd
> make that work for all service configurations.  That's my plan,
> anyway.  Configurations should, instead of complaining that provided
> types are bad, attempt to coerce first.
>
> I think the new interface will look more like:
>
> public interface ComponentClassTransformer
> {
>  void transformClass(PlasticClass componentClass, MutableComponentModel
> model);
> }
>
>
> On Tue, Feb 8, 2011 at 1:25 PM, Igor Drobiazko <igor.drobiazko@gmail.com>
> wrote:
> > Having an official road map would be a great improvement for Tapestry.
> Also
> > having two releases in 2011 is great.
> >
> > My plans for 5.3/5.4 are:
> >
> > - Support for JSR-330 (almost done)
> > - Support for some HTML5 features
> > - Support for "multidimensional" caching of pages, components, etc.
> > - Helping you out by migrating from javassist to plastic (I already
> started
> > learning ASM + Plastic)
> > - Maybe some support for other frameworks: Event though we having JPA
> > support in Tynomo, I'm thinking to move it into Tapestry? Such a widely
> used
> > framework should be supported.
> >
> > Along the way I'm working on the book which hopefully will be released
> > before summer.
> >
> > What do you think about moving some stuff from tapx to Tapestry core?
> >
> > On Tue, Feb 8, 2011 at 7:13 PM, Howard Lewis Ship <hlship@gmail.com>
> wrote:
> >
> >> Been chatting with clients (who may help fund this) and just thinking
> >> about plans.  Here's a rough outline of what I think I can commit to
> >> in 5.3 and 5.4.
> >>
> >> 5.3
> >> - Deprecate Javassist inside ComponentClassInstantiator, replace with
> >> Plastic
> >> - Deprecate ClassFactory, provide necessary hooks to use Plastic
> >> - Move Plastic into Tapestry?
> >> - Gradle build for Tapestry
> >> - Improve debugging experience (shadow per-thread values into shared
> >> object fields in development mode)
> >> - Improve asset pipelines for
> >>  - Dynamic generation of content (example, .less files converted to
> >> static CSS automatically)
> >>  - JS/CSS minimization
> >> - Do something about Component Report ... turn it into an Ant task,
> >> perhaps, or integrate Component Report into JavaDoc directly
> >> - Minor JS improvements, set expectations for 5.4 rewrite
> >>
> >> 5.4
> >> - Remove Javassist entirely
> >> - Remove ClassFactory
> >> - Rewrite JS entirely, introduce abstraction layer and backwards
> >> compatibility layer
> >> - Maybe cometd/server-push support
> >>
> >> I'd love to see both these releases in 2011.
> >>
> >> In case you missed in:  http://github.com/hlship/plastic
> >>
> >> --
> >> Howard M. Lewis Ship
> >>
> >> Creator of Apache Tapestry
> >>
> >> The source for Tapestry training, mentoring and support. Contact me to
> >> learn how I can get you up and productive in Tapestry fast!
> >>
> >> (971) 678-5210
> >> http://howardlewisship.com
> >>
> >> ---------------------------------------------------------------------
> >> To unsubscribe, e-mail: dev-unsubscribe@tapestry.apache.org
> >> For additional commands, e-mail: dev-help@tapestry.apache.org
> >>
> >>
> >
> >
> > --
> > Best regards,
> >
> > Igor Drobiazko
> > http://tapestry5.de
> >
>
>
>
> --
> Howard M. Lewis Ship
>
> Creator of Apache Tapestry
>
> The source for Tapestry training, mentoring and support. Contact me to
> learn how I can get you up and productive in Tapestry fast!
>
> (971) 678-5210
> http://howardlewisship.com
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: dev-unsubscribe@tapestry.apache.org
> For additional commands, e-mail: dev-help@tapestry.apache.org
>
>
>

Mime
  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message