Mailing-List: contact taglibs-dev-help@jakarta.apache.org; run by ezmlm
Date: Tue, 26 Dec 2000 09:51:41 -0500 (EST)
From: "Joseph B. Ottinger" <joeo@epesh.com>
To: taglibs-dev@jakarta.apache.org
Subject: RE: best practices (follow-up to the roadmap thread)
In-Reply-To: <NCBBKLKCHGCKLBAFEDFDMEJCHHAA.mcannon@internet.com>
Message-ID: <Pine.LNX.4.21.0012260938210.28369-100000@cupid.suninternet.com>
MIME-Version: 1.0
Content-Type: TEXT/PLAIN; charset=US-ASCII

On Wed, 27 Dec 2000, Mike Cannon-Brookes wrote:

> > I'm one of those CFML programmers, and stopped after scalability issues
> > was murdering a project. CFML's nice, and the tag generation is
> > okay... but I don't see that methodology lasting in Java. The cultural
> > dissonance is too great.
> 
> This is total BS IMHO. The tag generation is great in CF, granted it could
> be improved but so could JSPs.
> 
> I still see people trying to pigeonhole JSP programmers. If someone wants to
> use JUST the servlet 2.3 spec to write applications (and not J2EE - ie
> anyone using just Tomcat) they're writing applications in a very similar way
> to CF... therefore we should give them equivalent powered toolsets. If
> someone wants to go J2EE with EJBs etc, good for them. Doesn't mean everyone
> should be forced to go this way.

This is a valid point, and one I was waiting to see. I've been ignoring
this thought on purpose, because I *like* stereotyping programmers. Sue
me. I'm still thinking in terms of standards and what they bring, which is
where all this started: a standard JSP taglib. 

> > Ah! But an SQL resultset is not for your average HTML dummy. It implies
> > the person knows how to loop, knows about datatypes, knows about
> > databases, has access via a username/password, etc., possibly even more
> > than your average CFML programmer knows about the system configuration,
> > because CF configures data sources more transparently (currently, and
> > IMHO) than JSP does. An SQL taglib, and an XSL taglib, too, by their mere
> > existence are targeted for slightly more erudite people than you're
> > talking about.
> 
> Ahh, but your 'average dummy' isn't programming web apps. There is a modicum
> of intelligence required to even program HTML. Assuming you write a SQL
> taglib, they will need to know SQL too (otherwise everyone is screwed). If
> they know SQL, you can write a tag to loop through the results - and voila!
> Mr Dummy uses RecordSets without knowing them. I think people undersell CF
> programmers as 'base'. JSP can configure datasources in exactly the same way
> CF does (you only know the name). Same goes for XSL, 'mere web programmers'
> are learning XML/XSL, if we can make their job easier with a taglib, why the
> hell not? ;)

Because then we're out of a job. More seriously... I guess my primary
thought here is that if they're able to handle looping through a
resultset, etc., they're up enough on the thought processes involved that
they can use more "advanced" tags.

> > I wouldn't write my own SQL taglib. If I was going to, I'd have done it a
> > long, long time ago, in a galaxy far, far away - If I need db access, and
> > I do, I use EJBs or wrap the access up in a bean.
> 
> *shrug* I used CF for ages, I found the 'model' of accessing a database
> quite palatable. Certainly better than using SQL coded in BEANS (didn't you
> get taught anything Ep? ;)) which need to be recompiled. (You're spoilt by
> Orion's autocompilation of beans!) As for EJBs, they're fantastic but don't
> work in servlet containers.

I'm not spoiled by Orion's autocompilation - I'm old school, I still do it
myself. Two comments: "properties" and "I don't use mere servlet
containers." :)

> > > No argument.  I just think it should be easier and better than having to
> > > resort to nested tags because the technology wasn't thought out ahead
> > > far enough to provide a nice, clean expression syntax for tag
> > > attributes.
> > I'm trying to urge for orthogonality, so that once you know JSP's basic
> > (and current, for that matter) syntax, you don't have to keep working out
> > alternatives. ("You do it this way, except in this case, where you do it
> > THIS way, except in this other special case, where you can do it this way,
> > this way, or this way... and here you use a pointer." <-- exaggeration for
> > the point of illustration. You may remember what this is called. I don't.)
> 
> "waffle" ;)

Well, it's a form of pilpul, dunno the REAL term for it, you hoser.

> > > You're not being generous or sympatheitc enough in your reading of the
> > > idea (IMHO).  Like I said, Mike presented one thing, but the underlying
> > > idea was more generic: dynamic tag handler generation.
> > Hey, I'm not a generous or sympathetic person! And I know what Mike was
> > talking about; I also have some ideas as to how it could be addressed in
> > the real world. I'm playing the role of technocurmudgeon, remember? The
> > advocate of the priesthood, the guy who looks at handing fire to the
> > masses and says, "You know, people are going to manage to set their caves
> > on fire."
> JSPT is not geared towards generating content, it's geared towards
> generating HTML. And as I stressed MULTIPLE times, it's not ALL tags do, but
> it's a large part of what they should do. Not all HTML is included. At the
> moment tags that produce HTML are fugly, show me your solution. (That
> presumably is bounded by not having to compile, and having very little Java
> code if possible ... JSPT?)

Well, I didn't say it was your solution for ALL tags, and I don't consider
JSPT suitable for HTML only - it's suitable for CONTENT, like JSP itself
is.

-----------------------------------------------------------
Joseph B. Ottinger                           joeo@epesh.com
http://epesh.com/                             IT Consultant