• E-mail
• Print
• Comment
• Font Size
• Digg
• del.icio.us
• Discuss article

Searching for Desktop Integration

Posted on: Tuesday, 23 August 2005, 03:01 CDT

Integration Watch

The goal of all integration is access. Whether it is universal access to classrooms, as in Brown v. Board of Education of Topeka (Kan.), or universal access to data, the overarching goal is the same.

In IT, technologies such as EAI attain integration by normalizing data, ESBs do it by sneaking data transformations along the transport, and Web services do it by wrapping data transformations with verbose narratives. But how do you integrate data? Not your company-you personally. Chances are good that while you wrestle with the data silos in your enterprise, your desktop is nothing but silos.

Power users' desks have silos for e-mail, calendar data, documents, code, Web pages, software and so forth. Within these domain silos are smaller silos. Take e-mail and its standard organization mechanism, folders. Many times, a specific e-mail could be filed in any one of several folders for later retrieval. The problem is, I don't know which folder I'll think of when I want to retrieve it, and I don't want to copy the same data item into multiple folders. The exact same problem exists with files. Many times I have a file that could be placed in any of several directories. (Unix symbolic links are a partial solution here.)

The robust solution to breaking down silos, according to Google and several other firms, is search. For exam-' pie, Google Mail has no concept of folders. You can mark conversation as being associated with a given label, and any thread can have many different labels, so you can retrieve messages thematically. But all messages are stored in a single indivisible folder, and the principal retrieval mechanism is search. This metaphor works brilliantly. Searches are fast, and it's easy to zoom in on the e-mail you're looking for.

The trouble with Google Mail is that it works on mail only. (There are other limitations: Notably Google Mail does not search attachments.) What I need to do is search all my documents, e- mails, files, data. Enterprise search tools, such as the Google appliance or IBM's Information Integrator OmniFind, don't really do this. They're designed to search databases, Exchange containers and internal Web sites. They don't really do your desktop.

One very common desktop integration tool is a separate product from Google: the company's desktop search engine, which is available at desktop.google .com/business. This looks just like the online version of Google but for your desktop only. It is useful in integrating disparate types of documents in a single search; but after using it for a month, I must say my overall experience remains frustrating. There are several interesting limitations.

The first of these appears right after installation. The crawler starts crawling your database, making an index that will feed the search engine.

The trouble is that you have no specific control over this crawler. It runs when it wants to. If you just add a library of articles, Google Desktop solution will scan it whenever it decides to. Remove a file and who knows when that even will be recognized. Per the company, the only way to force a recrawl is to uninstall and reinstall the software. Not ideal on a system where files come and go or change location frequently.

In addition, the crawler is limited. As to e-mail, it reads only Outlook and Netscape mail. Eudora users are out of luck (which is odd since Eudora uses standard text files to hold e-mails). The system also skips important file types: Help files are not crawled (which is a serious limitation; often what I am looking for is in a help file), nor are .zip files, and PDF files are not crawled well. Moreover, there is a 5,000-word limit per document on all crawls. So, even if the software could read help files, you might still not find what you want even though the data is there.

It's hard to throw stones at free software, but it certainly seems that these limitations are intentional-Google could remove them. And now that I've seen what a desktop search engine can deliver, I am convinced that I would happily pay for the software if these limitations were lifted.

Unfortunately, open-source offerings are fairly thin. Of the few such projects, Zilverline project (www.zilverline.org) is the most advanced and mature. It addresses many of the concerns I have about Google's product. But it's not quite turnkey, as you have to install and configure Tomcat first, then load and configure Zilverline-but it's a good start.

Given the proliferation of data and documentation, productivity requires the integration of desktop through search. The tools, even if still not fully formed, are worth exploring now, and they will, I believe, become essential soon.

Andrew Binstock

Andrew Binstock is the principal analyst at Pacific Data Works.

Copyright BZ Media LLC Aug 15, 2005

Source: Software Development Times

More News in this Category