Friday, June 20, 2008

KnowledgeTree document download via Google Gadget

This is an installment feature for the KT-Gadget, a widget for KnowledgeTree, an open-source document management system.

As mentioned in the previous post, I employed an application proxy to act as a bridge between a Google Gadget and KnowledgeTree's SOAP services. Originally, the application proxy responses were formatted in XML. Moving forward, I decided to adopt the JSON format instead of XML owing to simpler methods in handling JSON feeds from both PHP and (of course) Javascript worlds.

The application proxy's SOAP client is based on PHP and converting an associative array of response items to JSON format is implemented simply by using PHP's json_encode function. At the gadget side, the requests for remote data are invoked using Google's _IG_FetchContent function. Using Javascript's eval() function, the returned JSON text response is converted to an object whose members can be easily accessed using dot or subscript operators.

Here are the details.. To download a document, KT-Gadget sends a request to the helper application by invoking the url:

http://kt-gadget.pipoltek.com/pipoltek.com/json/?oid=download_document&sid=rsop4e364ptf4slkp2a2vtvb96&did=4

The JSON response by the application proxy is of the form:
{"response":{"status_code":0,"message":"http:\/\/docs.pipoltek.com\/ktwebservice\/download.php?code=d2db2c0868ab0c9cd053fe2e43ab9926d1bcf94c&d=4&u=6a0j9qqpqc11de6nju6o0ej5o7"}}

The relevant block of KT-Gadget code is reproduced hereunder:

var url = "http://kt-gadget.pipoltek.com/pipoltek.com/json/?oid=get_folder_contents&sid=" + sessionID__MODULE_ID__ + "&fid=6";
_IG_FetchContent(url, function (jsonFeed) {
var jsonData = eval('(' + jsonFeed + ')');
if (jsonData.response.status_code != 0) {
alert ('Error in retrieving document!');
} else{
dd_link = jsonData.response.message;
_gel("ktg-lnk").innerHTML = '<a href="' + dd_link + '">@';
}
})

Once the application proxy has responded, the JSON text response (jsonFeed) is converted to an object (jsonData) using the eval() function. The response contains the download URL the value of which can be referenced as jsonData.response.message.

Where did the Session-ID (sid) and Document-ID (did) used as URL parameters come from? The Session-ID is assigned during login and the Document-ID, together with the rest of the document details, is returned by a call to get_folder_contents.

Here's the login invocation:
http://kt-gadget.pipoltek.com/pipoltek.com/json/?oid=login&uid=rexjun&pwd=cabanilla

The invocation to get_folder_contents:
http://kt-gadget.pipoltek.com/pipoltek.com/json/?oid=get_folder_contents&sid=0j8jkhjn55t7r3b0rshbcchpn1&fid=6

If you get a "Session is invalid" error, it's because the link is using an ID of an expired session. Change the URL parameter (sid) to a fresh Session-ID returned by the login call to get better results.

You may try out KT-Gadget by adding it to your iGoogle page or view the gadget source code here.

Thursday, June 12, 2008

Accessing Soap Services from Google Gadgets

The earlier gadget projects I developed have utilized third-party web services from a variety of sources:

  • Google Maps for geographic data
  • Google Docs spreadsheet as XML datasource
  • IP-Geolocation (IP address->Country) service from HostIP.Info
  • User profile and gaming statistics as XML feeds from Xfire.
Given the constraints of a space-restricted, remotely-hosted, javascript-only application, gadgets primarily rely on web services to provide richer data content without the attendant complexity. However, javascript is beset by cross-domain restrictions that even impact xmlHttpRequest, the very underpinnings of Ajax. Since Google gadgets are javascript apps, the same-origin security policy prevents it from making requests outside the domain it is running on.

Although Google addressed this problem by providing the _IG_FetchXMLContent function, it lacks support for https, SOAP and POST. It looks like the newer gadget.io.makeRequest function is intended to address these deficiencies. However, the availability of the gadget.io.* functions in Google's production servers is yet forthcoming.

I decided not to hold my breath..

KT-Gadget is a widget front-end to KnowledgeTree - an open-source document management system. To access KnowledgeTree's SOAP services, I had to deploy another application to receive the requests from Google's _IG_FetchXMLContent calls, make the corresponding SOAP requests, and generate the appropriate XML output. Two basic functions are implemented: login and get_folder_contents. The functions are based on PHP, the same language that was used to develop KnowlegeTree.

After pushing the login button, the gadget will indicate the progress of the subsequent operations:
  • Logging in
  • Fetching Documents
  • Sorting Documents
In its current state, the gadget is useful as a watcher of some sort of inbox for workflow-enabled documents like incoming faxes needing disposition or requisitions requiring approval. Sit around and watch your work pile up...

I plan to implement two more functions: download_document and perform_document_workflow_transition. We don't want work to pile up, do we? Other nice-to-have features: watch all accessible folders and sub-folders; show number of documents per folder; folder and document colors based on certain thresholds like document count, age, metadata contents (priority, etc); automatic refresh.

But don't hold your breath...

Thursday, May 22, 2008

Xfire Gadget

Have XML feed, will gadgetize..

Xfire provides xml feeds on gamer profiles, gameplays, friends, screenshots -- a whole lot of goodies. An Xfire gadget would be a good addition to one's blog, website or iGoogle.

One of the goals was to determine how close the gadget can resemble the real thing so it was initially developed using the default Xfire skin. Making a cross-browser scrollable table was not a trivial task and once the problem was solved, I started to hate the over-sized scroll bars. More coding lies ahead..

Employing the Ajax paradigm of single-page experience, remote services, rich visual controls and other design patterns, any web app can now come real close to the regular desktop UI.

A live gadget using a COD4 skin is displayed in another blog.

Tuesday, April 22, 2008

OpenID login for Knowledgetree

I have replaced (not integrated) KnowledgeTree authentication with OpenID. Supporting both default and OpenID authentication requires more work and that can come later.

This is by no means final since only limited testing has been performed so far. I've decided to post this in its raw form so I can solicit comments and suggestions from the community. And besides, my blog has not been updated for awhile now..

I'll detail the procedures here. I took notes but I may have missed something. Diligence is not one of my virtues..

  • Copy login.php to login-orig.php (save original, let's modify login.php)
  • Change template ktcore/login -> ktcore/openid
  • Copy ./templates/ktcore/login.smarty to openid.smarty
    - change invocation of stylesheet kt-login.css -> kt-openid.css
    - change the Username prompt label to OpenID
    - add class="openid" to input tag for username
    - change name="username" to name="openid_url"
    - remove the password input field

  • Download the OpenID logo
  • Upload logo as openid-bg.png to ./resources/graphics
  • Copy ./resources/css/kt-login.css to kt-openid.css
  • Add input.openid to kt-openid.css, as follows:
    input.openid {
    border: 1px solid #666;
    width: 232px;
    background: url(../graphics/openid-bg.png) no-repeat;
    padding-left: 18px;
    }
Note the OpenID logo in the input field. This is implemented by the input.openid section of the stylesheet. I also changed the input field name to openid_url in conformance to OpenID specifications. I'm using Verisign PIP as OpenID Identity Provider and the Seatbelt browser plug-in kicks in only when openid_url is used as identifier.

I installed version 2.x.x of the PHP OpenID Library, as follows:
  • Create the directory ./thirdparty/OpenID
  • Copy the file common.php found in the examples directory. Also copy the Auth directory from the library.
  • Add the following functions to common.php
    function fixslashes($s) {
    return get_magic_quotes_gpc() ? stripslashes($s) : $s;
    }

    function normOpenIDUrl($oid_url) {
    $claimed_id = strtolower (fixslashes($oid_url));
    $has_scheme = preg_match ('#^https\://#', $claimed_id) === 1;
    $has_scheme = $has_scheme || preg_match ('#^http\://#', $claimed_id) === 1;
    $has_tslash = preg_match ('#/$#', $claimed_id) === 1;
    return (($has_scheme?'':'http://') . $claimed_id . ($has_tslash?'':'/'));
    }
  • Overwrite the original functions in common.php with these
    function getReturnTo() {
    return sprintf("%s://%s:%s/login.php",
    getScheme(), $_SERVER['SERVER_NAME'],
    $_SERVER['SERVER_PORT']);
    }

    function getTrustRoot() {
    return sprintf("%s://%s:%s/",
    getScheme(), $_SERVER['SERVER_NAME'],
    $_SERVER['SERVER_PORT']);
    }
  • Change config/dmsDefaults.php to include the library in the path..
    $KTInit->prependPath(KT_DIR . '/thirdparty/OpenID');
The rest of the changes are in login.php, the source of which can be found here.

I've disabled automatic sign-up so make sure to create an OpenID account with Administrator privileges before you apply the changes. To prevent duplicate accounts, the usernames are OpenID URLs in normalized form (with leading http:// and trailing /). Thus, the username http://pipoltek.blogspot.com/ applies to any of the following acceptable OpenID identity URLs:
  • pipoltek.blogspot.com
  • http://pipoltek.blogspot.com
  • http://pipoltek.blogspot.com/
I've successfully authenticated using my OpenID accounts from Blogspot, Technorati and Verisign PIP. However, I had login failures using my Yahoo OpenID and a delegate identity URL.

My testbed:
  • VMWare Server version 1.0.4.56528
  • KnowledgeTree OSS 3.4.2 VM Appliance from rPath
  • Microsoft Windows XP Home Service Pack 2
  • Mozilla Firefox 2.0.0.14

Sunday, April 6, 2008

Verisign Labs PIP, my favorite OpenID Provider

As a long-time user of Yahoo, Blogger, and Technorati, I inherited the following OpenIDs:

  • pipoltek.blogspot.com
  • me.yahoo.com/rexjunjose
  • technorati.com/people/technorati/rexjun
In due time, I may have another one courtesy of Flickr. I enrolled at myopenid.com and that added rexjun.myopenid.com to my arsenal of OpenIDs.


My last OpenID provider might as well be the PIP, or Personal Identity Provider, from Verisign Labs. And this is primarily due to their Seatbelt Firefox plugin which adds ease and security when using your OpenID.

SeatBelt detects that you have clicked on an OpenID sign in field while not signed into your PIP account and prompts you to sign in. Once you have signed in, SeatBelt automatically returns you to the OpenID sign in page with your PIP URL filled in. The sign in session continues as normal.

Since redirection to the login page of your OpenID provider is done by the Seatbelt browser plugin, you can be assured that you are sending your credentials to Verisign and not a probable fake one generated by a phishing site.


One other neat feature is support for OpenID delegation allowing me to use pipoltek.com as my OpenID URI in place of the lengthy rexjun.pip.verisignlabs.com. This is made possible by adding two link tags inside the head section of my home page.


Wait.. there's more! You can protect your Verisign PIP account with two-factor authentication by the use of a security card, token, or a Sandisk U3 flash drive. Verisign PIP even supports the US$5 Paypal security key aside from its own.


As a lead player in identity protection services and a recent member of the OpenID Foundation, one can expect Verisign to continue providing innovative solutions supportive of the OpenID framework. So far, I'm not disappointed.