DECLARATIVE PUSH ON WEB
Mikko Pohja
Department of Media Technology, Helsinki University of Technology
P.O. Box 5400, FI-02015 HUT, Finland
Keywords:
Push technology, REX, SSE, Comet, XForms, XBL.
Abstract:
Push technology is an essential component of modern WWW applications. With the ability of sending relevant
information to users in reaction to new events, enables highly interactive applications on WWW. What has
been present in desktop applications for decades is only now coming online. Actually, the push is usually
emulated using the pull technology, since, with the HTTP protocol alone, it is not possible to make a real
push. A declarative approach widely used on WWW has not yet adopted for the push methods. In this paper
are reviewed four declarative methods for the push-update on WWW. The methods are defined by combining
existing and upcoming Web technologies. The scope of this paper is a targeting the update and modifying
the document on client-side. To evaluate the methods, a use case is designed and implemented with all the
methods.
1 INTRODUCTION
Modern WWW applications are based on information
push and asynchronous updates in addition to the tra-
ditional information pull. Examples of such applica-
tions include chats, stock tickers, news services etc.
The common factor for all these is a server’s ability to
push new information on to user’s screen whenever it
appears on the server. Often, the document is modi-
fied partially, i.e., only the changed parts are updated.
That has brought WWW applications closer to desk-
top applications.
Even though, there are already numerous of these
highly interactive WWW applications, the push tech-
nology is not really sophisticated at the moment. In
reality, the information push is emulated using infor-
mation pull by polling at constant time intervals or
keeping the connectionopen until next update is avail-
able. Some research-oriented push systems (Kanitkar
and Delis, 1998) and (Xin et al., 2005) and push-pull
combinations (Deolasee et al., 2001) are introduced
in the literature, but they are not widely adopted.
The server push is comprised of the following five
phases (cf. Figure 1): 1. Data changes on server, 2.
Application server is notified, 3. Update event is cre-
ated, 4. Event is delivered into the client, and 5. User
Interface (UI) is modified. The operations are inde-
pendent from each other. That is, they can be realized
by totally separate components. When a data changes
for instance in a data base, an application server must
notified. An example solution is discussed in (Bry
and Pˇatrˆanjan, 2005). The application server creates
the update event according to the change. The up-
date event, which is expected to be in XML format
in this paper, can be delivered for the user agent in
several ways. The delivery is independent from the
actual Document Object Model (DOM) modification
method. However, it is noteworthy that the traditional
WWW architecture using HTTP connections must be
augmented somehow to achieve the server push.
Figure 1: Server push phases.
This paper studies options to make the server push
for Web documents. The focus is in the final DOM
modification method and how it is defined, i.e., the
phase 5 in the Figure 1. The modification phase can
be further divided into two independent phases: tar-
geting the event to the correct element and the actual
UI modification. In the current systems, the both op-
erations are implemented by ECMAScript (ECMA,
1999). It provides diverse means to control those op-
erations. However, the trend on WWW has been to
define declarative technologies to replace the most
used ECMAScript functionality. That can be done
also for the push-updates. There are methods to de-
fine both the targeting and the UI modifications in a
declarative manner. In addition, some of them enable
an automatic evaluation of the content of an update.
201
Pohja M. (2008).
DECLARATIVE PUSH ON WEB.
In Proceedings of the Fourth International Conference on Web Information Systems and Technologies, pages 201-207
DOI: 10.5220/0001525002010207
Copyright
c
SciTePress
The push methods reviewed in this paper are com-
bined from existing Web technologies. The UI se-
mantics are provided by XHTML, CSS, XForms (Du-
binko et al., 2003), or XML Binding Language 2.0
(XBL) (Hickson, 2007) or combinations of them. Re-
mote Events for XML (REX) (Berjon, 2006) is used
for the targeting the updates. The methods are eval-
uated by implementing the use case described in the
next Section. The main contributions of this paper are
the following:
Four declarative push methods for WWW docu-
ments are defined.
A use case is designed and implemented with all
the four methods.
The declarative push methods are evaluated based
on the use case implementations.
The paper is organized as follows. After the use
case description, a background to the topic is given.
Sections 4 and 5 discuss the push-update methods and
the implementations, respectively. The methods are
evaluated in Section 6, a discussion is presented in
Section 7 and Section 8 concludes the paper.
2 USE CASE
Consider the following airline scenario as a use case
for push-updates. Joe is embarking on a business trip.
He is currently driving to the airport. His mobile
terminal beeps and he receives a message, which is
opened by the browser on his terminal. The message
notifies him that the plane is leaving on time and asks
if he wants to check in for the flight.
When Joe has confirmed the check-in, he receives
an electronic boarding pass. The boarding pass in-
cludes information on the flight, such as the seat, the
gate, and the boarding time. The boarding pass is a
Web document, which can receive updates about the
flight status.
At the airport, Joe’s mobile terminal beeps again.
The electronic boarding pass alerts him about the new
gate number. Later, the airline has published a new
departure time for his flight. Joe is proactively noti-
fied by the electronic boarding pass about the delay
and the new departure time. Finally, Joe is notified
that the boarding has started and that he should now
proceed to the gate.
Since the flight was delayed, Joe will miss his con-
necting flight to the final destination. The airline dis-
ruption management center is looking for an alterna-
tive routing for him. During the flight, Joe’s mobile
terminal is connected to the airplane’s WLAN net-
work. On the flight, he receives another notification.
It proposes a new routing for his trip. When he ac-
cepts the routing, he will receive a confirmation about
the traveling arrangements including the flight num-
bers and the departure and the arrival times.
The above scenario introduces an airline system,
which differs considerably from the legacy systems.
The fundamental difference is the ability to push in-
formation to the travelers’ Web browsers in their mo-
bile terminals. The scenario introduces two main
information types that are pushed for the travelers:
whole documents and piece of information that aug-
ments the existing document. Figure 2 gives examples
of both of those types in this scenario.
Figure 2: Push-updates in the airline scenario.
3 BACKGROUND
This paper studies options to make the server push for
Web documents. The focus is in the final DOM mod-
ification method and how it is defined. That is, the
whole document push mentioned in the previous Sec-
tion is out of scope. It could be realized by pushing
the URL of a document or the whole document. In
this Section is given a background what is needed to
realize the server push on WWW. In Figure 3 is de-
picted a legacy Web application architecture and the
additions it needs to support the server push. The
Eventing Service reacts on changes and notifies the
Notification Service. That, in turn, creates the events
and pushes them to the user agent either directly or
via Web Server. That is discussed in detail in next the
Subsection. Further below is explained the concept
how to target the DOM event for correct node in the
document to enable the modification.
3.1 Server Push
The HTTP connection provides information only
through a pull method. That is, a user agent asks for
the information from the server. Even if the content
changes, it is not delivered for the user agent unless it
asks for it. Server push is possible to implement with
WEBIST 2008 - International Conference on Web Information Systems and Technologies
202
Figure 3: Legacy Web application architecture augmented
with the server push components.
a privileged code and it can be emulated in many
ways.
At its simplest, Server push can be emulated by
polling the server at a certain time interval. That can
be done for instance by reloading the document or
with Ajax (Garrett, 2005) by asking incremental up-
dates. That causes a lot of extra net traffic, especially
the reloading, and there is always trade-off between
the latency and the polling frequency.
Comet (Russell, 2006) combines two existing
frameworks to emulate the push. It uses Ajax to real-
ize asynchronous updates and Multipart MIME type
(Levinson, 1997) for long-lived HTTP connections.
With Multipart, Ajax updates can be sent whenever
an event occurs on server-side. The downside is that
the server must keep connections open to all clients
it is to update. The central server must be able to
distribute the communication properly (Resig, 2006).
HTML 5 specification (Hickson and Hyatt, 2007)
defines an approach called Server Sent Events (SSE).
With SSE, an author can define a source, from
which the browser is listening for the incoming
connections. This technique requires privileged code
in a user agent and it is not widely implemented in
the current browsers. The HTML 5 specification
is at work-in-progress stage at World Wide Web
Consortium (W3C) at the moment.
Another real Server push is achieved using a
Server-Centric Interaction Architecture (SCIA). It
uses Session Initiation Protocol (SIP) to establish a
connection between the client and the server. The
client still fetches the Web documents through HTTP
but receives push-updates for the documents via SIP.
In this paper, SCIA is used to deliver the updates
for the client. SCIA was developed as a part of
WeSAHMI
1
project. An article about SCIA is being
prepared at the time of writing.
1
WeSAHMI Project,
http://www.tml.tkk.fi/Research/wesahmi/
3.2 Remote DOM Events
In addition to the delivery of the update, one needs
a method to target the event to a node and define the
modification. That can be done in a declarative man-
ner or via a scripting language. An XML update lan-
guage called XML-RL is represented in (Liu et al.,
2003). It treats the XML document as a data base
and provides methods to update it according to the
changes. However, it does not provide means for the
remote update. Another declarative proposal is REX.
W3C has produced a working draft of it, but at the
present moment has finished the specification work
because of patent issues
2
. In spite of all, REX repre-
sents the declarative concept to describe the event. In
this paper, that concept is called Remote DOM Events
(RDE).
3.3 XBL
XBL provides methods to enhance DOM nodes and
alter their sub trees. That is realized by attaching
functions, event handlers, style declarations, and sub
trees into the nodes. The complements are called
bindings. XBL is designed to augment the user ex-
perience of a document and not be a primary method
to describe the semantics. That is, the default UI se-
mantics must be defined by some other technology.
3.4 XForms
XForms 1.0 Recommendation is the next-generation
Web forms language, designed by the W3C. It solves
some of the problems found in the HTML forms by
separating the purpose from the presentation and us-
ing declarative mark-up to describe the most common
operations in form-based applications (Cardone et al.,
2005). It can use any XML grammar to describe the
content of the form (the instance data). Thus, it also
enables to create generic editors for different XML
grammars with XForms. It is possible to create com-
plex forms with XForms using declarative mark-up,
without resorting to scripting.
4 PUSH-UPDATE METHODS
The push-update methods consist of three operations,
which are mostly independent from each other and
even their order may vary depending on the used tech-
nologies. The operations and the implementation op-
tions are:
2
Report of the REX PAG, http://www.w3.org/2006/rex-
pag/rex-pag-report.html
DECLARATIVE PUSH ON WEB
203
Delivering the Update. An update must be pushed
to the client whenever it occurs on a system. Cur-
rent browsers support Comet-method; other pro-
posals include SSE and SCIA architecture.
Targeting the Update. When an update arrives on
client-side, there must be a way to target the new
UI mark-up for correct elements in the existing
document. In addition to RDE, that is possible
also with ECMAScript.
Adding the UI Semantics. By default, a system pro-
vides its updates in a raw data, which has to be
transformed into UI declarations. That can be
done on both server and client side. On server-
side, there must a dedicated component for that,
whereas on client-side the transformation can be
realized by CSS, XForms, XBL, or ECMAScript.
This paper focuses on the last two operations of
the push-update and expects that whatever the up-
date delivery method is, the payload is in XML for-
mat. Once the update is received, ECMAScript can
be used to target it to the correct element and give
the UI semantics. However, this paper focuses on the
declarative technologies. By combining the technolo-
gies mentioned above, we get the following options
to realize the push-update on the Web documents:
RDE combined with a User Interface Markup Lan-
guage (RDE+UIML), RDE+XForms, RDE+CSS, and
RDE+XBL. All these are discussed in the following
Subsections and summarized in Table 1.
Table 1: The push-update method summary.
Update Method Data Transform UI Semantics
RDE+UIML On Server UIML
RDE+XForms On Client XForms
RDE+CSS On Client CSS
RDE+XBL On Client CSS+XBL
4.1 RDE+UIML
In RDE+UIML, the data is transformed into a UIML
on server-side. That is, if the document to be updated
is, for instance, in XHTML or in Scalable Vector
Graphics (SVG) format, the content of the RDE event
is respectively a piece of XHTML or SVG mark-up,
too. In this method, there needs to be only RDE in-
terpreter on client-side, which targets the event to the
correct node. On server-side, there must be a compo-
nent which can transform the data into the UI mark-
up. It must also be aware which elements the mark-up
will replace or modify on client-side. That can be au-
tomated by retaining the client-side document also on
the server-side, even if it was created by a Servlet,
PHP, or any equivalent technology. When something
changes which affect to the document, an XML differ-
encing (XML diff) (Lindholm et al., 2006) is applied.
Its result is automatically in the client-side format and
can be sent to the client within RDE.
4.2 RDE+XForms
The RDE+XForms method is based on the XForms
technology, which enables to use the Model-View-
Controller paradigm in the Web forms. The benefit
is that the data of a form is separated from its presen-
tation. In other words, the data can exist in same form
in both the client and the server-side. The outfit of
the data is defined in an XForms form. To update the
XForms instance data with RDE events, the RDE in-
terpreter must update the instance document instead
of the UI document and call the refresh method for
the instance in the end. After the refresh, the update
appears on the UI. The relation of XForms instance
data with the actual UI DOM is represented in Figure
4. As can be seen, the instance is completely separate
from the UI. There is a two-way connection between
an instance node and its UI node.
Figure 4: The XForms instance in relation to the UI docu-
ment and the form constraints.
XForms can be used also to automatically evaluate
the content of the update. The XForms bind element
can retrieve the content and make calculations on it
or compare its value to other values in the instance.
The result can be made visible on the UI, too. Fig-
ure 4 depicts also a dependency graph through which
the XForms engine control the values of the data in-
stance. The bind elements affect to the graph among
the others.
4.3 RDE+CSS
RDE+CSS is used in a similar way as RDE+XForms.
Using CSS, the whole document on the client-side or
part of it is a sort of data instance. CSS role is to
define UI semantics of the data elements. It cannot
provide automatic evaluation of an update.
WEBIST 2008 - International Conference on Web Information Systems and Technologies
204
4.4 RDE+XBL
RDE+XBL is an extension to the RDE+CSS method.
XBL can be used to define the UI semantics a lot more
diverse than CSS. With XBL, the data instance can be
transformed into the desired UIML (e.g., XHTML or
SVG). Nevertheless, to not to abuse XBL, it cannot be
used to primarily define the semantics of the elements
in a user agent. That is why CSS is used to describe
the default semantics of the data instance in this paper.
In addition to the UI semantics, XBL can define func-
tions for the updated element. The functions can eval-
uate and react to the updated content like the XForms
bind element. To react to the updated data instance
element, an xblEnteredDocument() function must de-
fined within the XBL implementation element.
5 USE CASE
IMPLEMENTATIONS
The push-update methods were tested by implement-
ing the use case described in Section 2. SCIA ar-
chitecture was used to create the updates and notify
the application server. In Figure 3, SCIA locates in
the application server. An open source XML browser
called X-Smiles (Vuorimaa et al., 2002) was used as
a user agent on the client-side. REX was used as
the RDE technology and it was implemented for X-
Smiles. In this Section is described as an example
how a new departure time of a flight can be updated
with the push-update methods. The main differences
of the methods are discussed.
In RDE+UIML, the update’s UI semantics is de-
fined on server-side. In the use case implementation,
the UI description was in XHTML. Below is a sample
event which replaces a span element in the document.
The element is recognized by an id attribute. The
REX interpreter in a user agent automatically does the
operation defined in event element when it receives
the event.
<rex xmlns=’http://www.w3.org/ns/rex#’>
<event target=’id("edt")’
name=’DOMNodeRemoved’>
<span id=’edt’>16:15</span>
</event>
</rex>
To highlight the changed departure time, text
“DELAYED” is added after the updated time. That
was targeted to a different element. The REX event
was as follows:
<rex xmlns=’http://www.w3.org/ns/rex#’>
<event target=’id("edtmsg")’
name=’DOMNodeRemoved’>
<span id=’edtmsg’>DELAYED</span>
</event>
</rex>
The boarding pass after the updates is depicted in
Figure 5.
Figure 5: The boarding pass after the push-update.
In the other methods, the UI semantics is given on
the client-side. Thus, the REX event can be a bit sim-
pler. Below is an event which updates a data instance
on the client-side. In this case, the element’s name
describes the data it contains.
<rex xmlns=’http://www.w3.org/ns/rex#’>
<event target=’//edt’ name=’DOMNodeRemoved’>
<edt>16:15</edt>
</event>
</rex>
In RDE+XForms, the edt element resided within
the XForms instance element. The REX interpreter
must be able to notice that and updated instance doc-
ument instead of the UI document. After the up-
date, the instance document had to be refreshed to
update the UI. The XForms bind element compared
the updated time with the scheduled departure time
using if() function and set an edtmsg element’s con-
tent as “DELAYED” if departure time was later than
scheduled. The edtmsg element was referenced by
another output element in the form. In contrast to
RDE+UIML, there was no need for two update events
because the “DELAYED” message could be produced
by the XForms bind element. The final output was ex-
actly similar to the RDE+UIML implementation (cf.
Figure 5).
The RDE+CSS method requires that there is a
CSS declaration for the edt element to visualize the
data update. In the update like the one above, that is
usually the case because there already was another edt
element, which was replaced. In this case, the CSS
declaration was:
edt { display: inline; }
CSS does not provide methods to evaluate the
content of the update like XForms does. Thus, the
“DELAYED” message must be sent separately as
DECLARATIVE PUSH ON WEB
205
with the RDE+UIML method above. That could be
avoided using the RDE+XBL method. An XBL bind-
ing bound to an edt element could define a function
which checked if the new time differed from sched-
uled departure time and added the message if it did.
The binding is depicted below.
<xbl xmlns="http://www.w3.org/ns/xbl">
<binding element="edt">
<implementation>
({
xblEnteredDocument: function () {
var edt = this.boundElement.firstChild
.nodeValue;
var sdt = document
.getElementsByTagName(’sdt’)
.item(0).firstChild.nodeValue;
if (edt != sdt) {
var msg = document
.createTextNode(’DELAYED’);
var edtmsg = document
.getElementsByTagName(’edtmsg’)
.item(0);
edtmsg.replaceChild(msg,
edtmsg.firstChild);
}},})
</implementation>
</binding>
</xbl>
6 EVALUATION
The push-update methods are compared to each other
against the following criteria.
Maintainability of the UI Description. What have
to be taken into account, if an author wants to
modify the UI afterwards?
Automatic Evaluation of the Update. How the
methods can react on the content of the update?
Versatility. What constraints the methods have re-
garding the UI description formats.
Maintainability of UI description is simpler with
the methods in which UI description resides only on
the client-side. In RDE+UIML, the update’s UI de-
scription is defined on server-side, which requires
making modifications on two sites. However, it is
noteworthy, that if UIML is used with CSS, the UI
can partially be modified through CSS which does not
affect on the UI mark-up. Another way to avoid the
modifications on two sites with RDE+UIML is to use
XML diff (cf. Section 4.1) on the server-side.
As discussed in previous Section, the
RDE+XForms and RDE+XBL methods provide
means to react to the content of the update. With
other methods, extra updates and possibly server-side
logic are needed to realize the same functionality.
The RDE+UIML and RDE+XBL can be used
with any UI description language as long as it is in
XML format. The same goes with RDE+XForms
presuming that the format can be combined with
XForms. RDE+CSS can use only CSS to layout the
data.
7 DISCUSSION
The RDE+XForms and the RDE+XBL methods sur-
vived best from the evaluation above. They both are
easy to maintain, provides automatic evaluation of
the content of the updates, and work well with other
UI technologies. The RDE+XBL offers more diverse
evaluation of the content of the update than XForms
since it is defined by ECMAScript. That, on the other
hand, is against the declarative approach used in this
paper.
RDE+UIML loses on the maintainability and on
the content evaluation for the two above. However, it
provides the updates in the final format, which may
make the user agent simpler. The UIML can be basi-
cally anything supported by the user agent.
The RDE+CSS method’s power is its simplicity. It
is limited on CSS’s ability to define UI, but uses a data
instance on the client-side, which makes it simple to
use.
A notable aspect on the push-updates is that they
suit well only for UI elements which do not allow user
input. That is, for instance, if the XForms input el-
ement could receive updates, there should be logic
to prevent the update override a possible user input.
User may have provided her input but not sent it to the
server when the update arrives. For example, a movie
theater can offer a service where user can select a seat
from a seat map. The seat map shows both free and
reserved seats. All the users see the same map and the
reservations are updated for everyone. There must be
a way to avoid double booking, for instance by lock-
ing few seats for a short period for each user etc.
There is a huge amount of spam sent everyday by
e-mail. Undoubtedly, spammers are willing to use
also push-updates to deliver their messages. That can
be prevented by allowing the updates only from the
same source as the original document. In addition, it
might be necessary to ask a user if he wants to get
updates from a certain source. That way he can for
instance deny advertisements but accept information
updates.
WEBIST 2008 - International Conference on Web Information Systems and Technologies
206
8 CONCLUSIONS
In this paper, four declarative push-update methods
were studied, namely RDE+UIML, RDE+XForms,
RDE+XBL, and RDE+CSS. The methods were de-
fined combining existing and upcoming Web tech-
nologies. To evaluate the methods, a use case, which
included a lot of updates, was designed and it was
implemented with all the four methods. The imple-
mentations and the implementation experience were
reported in this paper.
It was easiest to implement the use case with
RDE+XForms and RDE+XBL. They both provide a
data instance on the client-side and a powerful UI
control. They can also be combined with other Web
technologies easily. RDE+UIML and RDE+CSS are
simpler methods, but also with them it was possible to
implement the use case. However, they required more
update messages to complete the same task. Anyway,
even the RDE+CSS method proved to be useful in
simple use cases.
Along with the adoption of these methods, the
WWW still needs a proper method to deliver the up-
dates. Nowadays, the push is emulated by the Comet
technology or even by polling the server. SSE seems
to be a promising technology for delivery, but com-
parison of delivery technologies is left as a future
work. Another item to work further is the updates of
the input elements. As discussed in the previous Sec-
tion, there must some kind of control if several users
can update a single data. The current solution fits only
for centralized updates.
REFERENCES
Berjon, R. (2006). Remote Events for XML (REX) 1.0.
Working draft, W3C.
Bry, F. and Pˇatrˆanjan, P.-L. (2005). Reactivity on the web:
paradigms and applications of the language XChange.
In SAC ’05: Proceedings of the 2005 ACM symposium
on Applied computing, pages 1645–1649, New York,
NY, USA. ACM.
Cardone, R., Soroker, D., and Tiwari, A. (2005). Using
XForms to simplify web programming. In WWW
’05: Proceedings of the 14th international conference
on World Wide Web, pages 215–224, New York, NY,
USA. ACM Press.
Deolasee, P., Katkar, A., Panchbudhe, A., Ramamritham,
K., and Shenoy, P. (2001). Adaptive push-pull:
disseminating dynamic web data. In WWW ’01:
Proceedings of the 10th international conference on
World Wide Web, pages 265–274, New York, NY,
USA. ACM.
Dubinko, M., Klotz, L. L., Merrick, R., and Raman, T. V.
(2003). XForms 1.0. W3C Recommendation.
ECMA (1999). Standard ECMA-262 - ECMAScript Lan-
guage Specification. Standard, ECMA.
Garrett, J. J. (2005). Ajax: A New Approach to Web Appli-
cations. Technical report, Adaptive Path.
Hickson, I. (2007). XML Binding Language (XBL) 2.0.
Candidate Recommendation, W3C.
Hickson, I. and Hyatt, D. (2007). HTML 5. Editor’s draft,
W3C.
Kanitkar, V. and Delis, A. (1998). Real-Time Client-Server
Push Strategies: Specification and Evaluation. In
RTAS ’98: Proceedings of the Fourth IEEE Real-Time
Technology and Applications Symposium, pages 179–
188, Washington, DC, USA. IEEE Computer Society.
Levinson, E. (1997). The MIME Multipart/Related
Content-type. RFC, RFC Editor, United States.
Lindholm, T., Kangasharju, J., and Tarkoma, S. (2006). Fast
and simple XML tree differencing by sequence align-
ment. In DocEng ’06: Proceedings of the 2006 ACM
symposium on Document engineering, pages 75–84,
New York, NY, USA. ACM Press.
Liu, M., Lu, L., and Wang, G. (2003). A Declarative
XML-RL Update Language. In 22nd International
Conference on Conceptual Modeling, ER’03, volume
2813/2003 of Lecture Notes in Computer Science,
pages 506–519. Springer-Verlag.
Resig, J. (2006). Pro JavaScript
TM
Techniques, chapter 14,
pages 287–304. Springer-Verlag, New York, NY,
USA.
Russell, A. (2006). Comet: Low Latency Data
for the Browser. Weblog. Available online:
http://alex.dojotoolkit.org/?p=545.
Vuorimaa, P., Ropponen, T., von Knorring, N., and
Honkala, M. (2002). A Java based XML browser for
consumer devices. In 17th ACM Symposium on Ap-
plied Computing, pages 1094–1099, Madrid, Spain.
Xin, Z., Zhao, J., Chi, C., and Sun, J. (2005). Informa-
tion Push-Delivery for User-Centered and Personal-
ized Service. In Second International Conference on
Fuzzy Systems and Knowledge Discovery, FSKD’05,
volume 3613/2005 of Lecture Notes in Computer Sci-
ence, pages 594–602. Springer-Verlag.
DECLARATIVE PUSH ON WEB
207