cocoon-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Stefano Mazzocchi <stef...@apache.org>
Subject [RT] Cocoon Emotional Landscapes
Date Wed, 11 Oct 2000 21:22:11 GMT
I was asked to write a white paper on why Cocoon is better than other
web technologies. This is mostly targetted to managers and executives.
Since it might help you to convince your boss, I'm posting it to this
list.

Any volunteer that translates this into the Cocoon Document DTD is
welcome :)

Also some spell checking and grammar correction.

Yes, you're right, I'm a lazy butt. :)

This is included into the [RT] (random thoughts) series because it
includes many references to things that are yet to come, either in this
project or on the web as a general thing. Moreover, RT's are more fun to
write :)

----------------------------- cut here ------------------------------


                     Cocoon Emotional Landscapes
                                or 
                 why you can't afford to miss Cocoon

                                by
                          Stefano Mazzocchi
                        (stefano@apache.org)



Introduction
------------

Everybody talks about XML. XML here, XML there. All application servers
support XML, everybody wants to do B2B using XML, web services using
XML, even databases using XML.

Should you care about it? Given the amount of hype, you can't afford to
go around ignoring the argument, would be like ignoring the world wide
web 10 years ago: a clear mistake. But why is this so for XML? What is
this "magic" that XML seems to have to solve my problems? Isn't this
another hype to change once again the IT infrastructure that you spent
so much time implementing and fixing in the last few years? Isn't
another way to spill money out of your pockets?

If you ever asked yourself one of the above questions, this paper is for
you. You won't find singing-and-dancing marketing crap, you won't find
boring and useless feature lists, you won't find the usual acronym
bombing or those good looking vaporware schemas that connect your
databases to your coffee machines via CORBA or stuff like that.

I'll explain you what the Apache Cocoon project is about and what we are
doing to solve the problems that we encountered in our web engineering
experiences, but from an executive perspective, yes, because we all had
the problems of managing a web site, dealing with our colleages, rushing
to the graphical guru to have the little GIF with the new title, or
calling the web administrator at night because the database is returning
errors without reasons.

I was frustrated of seeing the best and most clever information
technology ever invented (the web) ruined by the lack of engineering
practices, tortured by those "let's-reinvent-the-wheel-once-again"
craftmen that were great at doing their jobs as inviduals but that
couldn't scale and imposed a growth saturation to the whole project.

There had to be a better way of doing things.


personal experience
-------------------

In 1998, I volunteered to create the documentation infrastructure for
the java.apache.org project, which is composed by a bunch of different
codebases, maintained by a bunch of different people, with different
skills, different geographical locations and different degree of will
and time to dedicate to the documentation effort.

I designed the site look (which still remains the same), I designed the
guidelines for HTML creation (which had to run on *all* HTML browsers
that ever existed: tell that to your graphic designers and look at their
faces!), I wrote the scripts to regenerate the site automatically taking
documentation out of each project, and kicked butts around if they
weren't following the guidelines.

But pretty soon I realized no matter how great and well designed the
system was, HTML was a problem: it was *not* designed for those kind of
things. I looked at the main page (http://java.apache.org/) from the
browser and you could clearly identify the areas of the screen: sidebar,
topbar, news, status. But if you opened the HTML, boom: a nightmare or
table tags and nesting and small little tricks to make the HTML appear
the same on every browser.

So I looked around for alternative technologies, but *all* of them were
trying to add more complexity at the GUI level (Microsoft Frontpage,
Macromedia Dreamweaver, Adobe GoLive, etc...) hoping to "hide" the
design problems of HTML under a think layer of WYSIWYG looks.

What you see is what you get.

But what you see is all you've got.

How can you tell your web server to "extract" the information from the
sitebar? How can you have the news feeds out of a complex HTML page?

Damn, it's easy for a human reader: just look at the page and it's very
easy to distinguish between a sidebar, a banner, a news and a stock
quote. Why is it so hard for a machine?


The HTML model
--------------

HTML is a language that tells your browser how to "draw" things on its
window. An image here, a letter there, a color down here. Nothing more.
The browser doesn't have the "higher level" notion of "sidebar": it
lacks the ability to perform "semantic analysis" on the HTML content.

Semantic analysis? Yeah, it's the kind of thing the human brain is
simply great at doing, while computer programs simply suck big time.

So, with HTML, we went a step up and created a highly visual and
appealing web of HTML content, but we went two steps back by removing
all the higher level semantic information from the content itself.

Ok, let's make an example... I think most of you have seen an HTML
page... if not, here is an example:

 <html>
  <body>
   <p>Hi, I'm an HTML page</p>
   <p align="center">Written by Stefano</p>
  </body>
 </html>

which says to the browser:

 - I'm a HTML page
 - I have a body
 - I have a paragraph
 - I contain the sentence "Hi, I'm an HTML page."
 - I contain the sentence "Written by Stefano"

Suppose you are a chinese guy that doesn't understand our alphabet, try
to answer the following question:

 who wrote the page?

You can't perform semantic analysis, you are as blind as a web browser.
The only thing you can do is draw it on the screen since this is what
you were programmed to do. In other words, your semantic capacity is
fixed to the drawing capabilities and a few other things (like linking),
thus limited.


Semantic markup
---------------

Suppose I send you this page:

 <page>
  <author>sflkjoiuer</author>
  <content>
   <para>sofikdjflksj</para>
  </content>
 </page>

now, who wrote the page? easy, you say, "sflkjoiuer" did. Good, but now
I send you

 <dlkj>
  <ruijfl>sofikdjflksj</ruijfl>
  <wijlkjf>
    <oamkfkj>sflkjoiuer</oamkfkj>
  </wijlkjf>
 </dlkj>

tell me, who wrote the page? You could guess by comparing the structure,
but how do you know the two stuctures reflect the same semantic
information?

The above two pages are both XML documents.

Are they going to help you? Are they doing to simplify your work? Are
they going to simplify your problems?

At this point, clearly not so, rather the opposite.

So, I'm sure you are wondering, why did I spend the last two years
writing an XML publishing framework? At the end of this paper, you'll be
able to answer this question alone, so let's keep going.


The XML Language
----------------

XML is most of the times referred to as the "eXtensible Markup Language"
specification. A fairly small yet complex specification that indicates
how to write languages. It's a syntax. Nothing fancy at all. So

 <hello></hello>

is correct, while

 <hello></hi>

is not, but

 <hello><hi/></hello>

is correct. That's more or less it.

XML is the ASCII for the new millenium, it's a step forward from ASCII
or UNICODE (the international extention to ASCII that includes all
characters from all modern languages), it defined a "lingua franca" for
textual languages.

Ok, great, so now instead of having one uniform language with visual
semantics (HTML) we have a babel of languages each with its own
semantics. How this can possibly help me?


XML transformations
-------------------

This was the point where I was more or less two years ago for
java.apache.org: I could use XML and define my own semantics with
<sidebar>, <news>, <status> and all that and I'm sure people would have
found those XML documents much easier to write (since the XML syntax is
very similar to the HTML one and very user friendly)... but I would have
moved from "all browsers" to "no browser".

And having a documentation that nobody can browse is totally useless.

The turning point was the creation of the XSL specification which
included a way to "transform" an XML page into something else. (it's
more complex than this, but I'll skip the technical details).

So now you have:

 XML page ---(transformation)--> HTML page
                    ^
                    |
          transformation rules

that allowed me to write my pages in XML, create my "graphics" as
tranformation rules and generate HTML pages on the fly directly from my
web server.

Cocoon 1.0 did exactly this.


The model evolves
-----------------

If XML is a lingua franca, it means that XML software can work on almost
anything without caring about what it is. So, if a cell phone requests
the page, Cocoon just has to change transformation rules and send the
WAP page to the phone. Or, if you want a nice PDF to printout your
monthly report, you change the transformation rules and Cocoon creates
the PDF for you, or the VRML, or the VoiceML, or your own proprietary
B2B markup.

Anything without changing the basic architecture that is simply based on
the simple "angle braket" XML syntax.


Separation of Concerns (SoC)
----------------------------

Cocoon was not the first product to perform server side XML
transformations, nor will be the last one (in a few years, these
solutions will be the rule rather than the exception). What is the
"plus" that the Cocoon project adds?

I believe the most important Cocoon feature is SoC-based design.

SoC is something that you've always been aware of: not everybody is
equal, not everybody performs the same job with the same ability.

It can be observed that separating people with common skills in
different working groups increases productivity and reduces management
costs, but only if the groups do not overlap and have clear "contracts"
that define their operativity and their concerns.

For a web publishing system, the Cocoon project uses what we call the
"pyramid of contacts" which outlines four major concern areas and five
contracts between them. Here is the picture:

                               management
                              /    |    \
                            /      |      \
                        style - content - logic

Cocoon is "engineered" to provide you a way to isolate these four
concern areas using just those 5 contracts, removing the contract
between style and logic that has been bugging web site development since
the beginning of the web.

Why? because programmers and graphic people have very different skills
and work habits... so, instead of creating GUIs to hide the things that
can be harmful (like graphic to programmers or logic to designers),
Cocoon allows you to separate the things into different files, allowing
you to "seal" your working groups into separate virtual rooms connected
with the other rooms only by those "pipes" (the contracts), that you
give them from the management area.

Let's have an example:

 <page>
  <content>
   <para>Today is <dynamic:today/></para>
  </content>
 </page>

is written by the content writers and you give them the "contract" that
states that the tag <dynamic:today/> prints out the time of the day when
included in the page. Content writers don't care (nor should) about what
language has been used for that, nor they can mess up with the
programming logic that generates the content since it's stored in
another part of the system they don't have access to.

So <dynamic:today/> is the "logic - content" contract.

At the same time, the structure of the page is given as a contract to
the graphic designers who have to come up with the transformation rules
that transform this structure in a language that the browser can
understand (HTML, for example).

So, the page structure is the "content - style" contract.

As long as these contract don't change, the three areas can work in a
completely parallel way without saturating the human resources used to
manage them: costs decrease because time to market is reduced and
maintenance costs is decreased because errors do not propagate out of
the concern areas.

For example, you can tell your designers to come up with a "Xmas look"
for your web site, without even telling the other people: just switch
the XMas transformation rules at XMas morning and you're done.... just
imagine how painful it would be to do this on your web site day.

With the Cocoon architecture it's a couple of line changes away.


The Cocoon innovations
----------------------

The technologies defined in the XML model are the base of everything,
but many technologies and solutions were designed and invented by the
Cocoon project:

 - the XSP language (eXtensible Server Pages)
 - the sitemap concept
 - the resource view concept


The XSP language
----------------

There were no technologies that allowed us to create the kind of
"content - logic" separation outlined above, so we specified it. We used
the JSP page compilation model as a starting point, but we designed XSP
to be totally abstracted from programming language used.

This means that you can implement the tags using your favorite language,
without having to force your programmers to use a specific programming
language. At the time of writing, only the Java programming language is
implemented (being Cocoon written in Java), but it's easy to picture
development of other language hooks for XSP once Cocoon receives more
attention.

Anyway, thanks to complete separation of concerns, the changes in the
logic concern area don't impact the other areas and allow more
flexibility in the human resource used in the project since programming
languages will be easily interchanged and XSP is not connected to a
particular implementation (the AxKit project implemented XSP directly in
Perl)


The Sitemap concept
-------------------

XSP allowed us to complete the second contract, but we have three more
to go and all of them start from the site management area.

Site managers are responsible for controlling the resources that are
hosted on the site. A "resource" is not a file stored on a file system
but the view of the data connected to a particular web address.

For example, 

 http://your.intranet.com/calendar/today

is a "resource" but could be dynamically generated by information stored
on your corporate database.

The sitemap allows site administrators to "compose" the modules that
collaborate to create a particular resource and connect them to the
particular resource address. For example,

 <match pattern="/calendar/today">
  <generate type="calendar" src="today"/>
  <transform src="transformation/calendar2html"/>
  <serialize type="html"/>
 </match>

or, supposing you want to connect with your Palm to the calendar or with
Acrobat for printing:

 <match pattern="/calendar/today">
  <generate type="calendar" src="today"/>
  <select type="browser">
   <when test="palm">
    <transform src="transformation/calendar2palm"/>
    <serialize type="palm"/>
   </when>
   <when test="acrobat">
    <transform src="transformation/calendar2pdf"/>
    <serialize type="pdf"/>
   </when>
   <otherwise>
    <transform src="transformation/calendar2html"/>
    <serialize type="html"/>
   </otherwise>
  </select>
 </match>

yes, it's that simple, the above is a real working example. Given the
right logic components and transformation rules, the site administrators
to just that: compose the modules to create web sites and web
applications.

Of course, there are situations where the logic involved in the
componentization is much more complex than this, but the sitemap concept
allows to finish the implementation of the five contracts without
overlapping between the concern areas.


The Resource View concept
-------------------------

The third big innovation of the Cocoon project is the notion of
"resource views". It's kind of an abstract concept so I'd like to start
with an example to explain the problem.

A big XML myth is that a web of XML content would be easier to searching
and index given the uniform syntax.

This is simply dead wrong.

As one of the first examples clearly showed, there is no way for a
search engine to perform semantic analysis on something like

 <dlkj>
  <ruijfl>sofikdjflksj</ruijfl>
  <wijlkjf>
    <oamkfkj>sflkjoiuer</oamkfkj>
  </wijlkjf>
 </dlkj>

the only useful thing would be to ask for all the web resources where
"sofikdjflksj" is included into "ruijfl" tags... but if you think at the
possible huge babel of XML languages that might appear in the next
decades, this is a clear problem.

The RDF (resource description framework) model is an effort that aims to
reduce this problem by creating uniform semantics to "describe" any kind
of information. For example, it could allow the author of this weird
markup to indicate that

 <ruijfl>

inherits the semantic meanings of <author>. So, if you search for
"sofikdjflksj" as an <author>, it will be able to search between
<ruijfl> as well.

It will be a happy day when there will be enough RDF information on the
web to allow such semantic-rich searching, it will correctly be a much
better web for search and retrieval of information.

The problem is that all other publishing systems except Cocoon "hide"
this information inside the system, there is no standard way to "ask"
for the original RDF-ized semantic content of the requested resource.

There is no way to ask for a particular "view" of the resource.

In Cocoon you can define "views" for each resource or group of
resources: you can ask for the "content" view, or for the "schema" view
(that returns you the structure of the document and the information to
validate it), the "link" view that returns you the pages that are
connected to this particular resource and so on.

Resource views are a particular type of HTTP variants, but targetted for
the yet-to-be-possible semantic indexing of content.

Cocoon itself uses the view for its batch mode: it performs as a crawler
and saves a snapshot of the site on disk, useful for creating offline
documentation or CD-ROM snapshots of dynamic web sites.


Cocoon present
--------------

The Cocoon project is currently discussing new features such as "content
aggregation" that would simplify the creation of portal-like sites where
content is aggregated from different sources into the same page.

At the time of writing, the design has yet to solidity.


Cocoon future
-------------

In the future, Cocoon will provide local semantic searching capabilities
allowing you to gain immediate advantage of the time invested in
creating highly semantic content for your site.

I believe this is the only way to convince people to invest time and
resources into creating a better content model for their local
information. We still don't have any idea on how this will happen or how
it will work, but I believe the Cocoon project has a major role in the
promotion of the next web generation and semantic searching is a big
part of it.

We'll do whatever it takes to help the semantic web to exist.

A further future goal is to allow Cocoon to exchange semantic indexing
information in a Peer2Peer way to create a decentralized semantic search
engine... (even if there are big protocol scalability problems to
solve). Consider this high steam vaporware, but I believe that it will
make perfect sense to decentralize information searching capabilities to
avoid both central point of failure and central points of censorship in
the future.

Final words
-----------

If you reached this far by reading all sections, I was successful in
getting your attention and I think you are able to both understand the
importance of the Cocoon Project and distinguish most of the marketing
hype that surrounds XML and friends.

Just like you shouldn't care if somebody offers you a software that is
"ASCII compliant" or "ASCII based", you shouldn't care about "XML
compliant" or "XML based": it doesn't mean anything.

Cocoon uses XML as a core piece of its framework, but improves the model
to give you the tools you need and is designed to be flexible enough to
follow your needs as well as paradigm shifts that will happen in the
future.

Bibliography
------------

 W3C - Technical Reports - http://www.w3.org/TR/
 Tim Berners-Lee - Web Design Issues - http://www.w3.org/DesignIssues/
 Tim Berners-Lee - Good URIs don't change -
http://www.w3.org/Provider/Style/URI
 Jakob Nielsen - URL as UI - http://www.useit.com/alertbox/990321.html
 Roy Fielding at alt. - HTTP/1.1 (RFC 2616) -
ftp://ftp.isi.edu/in-notes/rfc2616.txt

Legal
-----

Copyright (c) 2000 Stefano Mazzocchi. All rights reserved.

Permission to distribute this paper in any form is granted provided it
is not modified in any part. For further permissions contact the author.

----------------------------- cut here ------------------------------

-- 
Stefano Mazzocchi      One must still have chaos in oneself to be
                          able to give birth to a dancing star.
<stefano@apache.org>                             Friedrich Nietzsche
--------------------------------------------------------------------
 Missed us in Orlando? Make it up with ApacheCON Europe in London!
------------------------- http://ApacheCon.Com ---------------------

Mime
View raw message