THUNDERSTONE NEWS

Spring 2005 - Archive

CONTENTS

• Appliance Small Business Edition
• New Search Appliance Features
• Customer Spotlight: Alibris Inc.
• Macintosh Versions
• Appliance on GSA Schedule
• Recent Reviews
• Meet Us in Washington April 5-7
• Summer Training
• Subscription/Unsubscription and Contacts

A new version of the Search Appliance is aimed at the small business market. Called Appliance SBE (Small Business Edition), it provides a search of up to 50,000 documents or web pages. The Appliance SBE contains most features of Thunderstone's enterprise-level appliance versions.

Thunderstone Search Appliances are unique in the appliance market in allowing customers to index third-party information, such as government or industry web sites. This policy enables customers to create unique targeted collections of web information for their users or for competitive intelligence.

We've been busy working our way through the long list of product enhancement suggestions you have sent in! Here are some of the more notable new features.

Metasearch

A common request by Search Appliance customers is to be able to search two or more "profiles" at once. Until now, that required some work to set up, using web services to submit multiple queries. No longer!

A profile is our name for what many users call an index or search collection. There are many reasons for maintaining multiple profiles. For example, they can be made available separately to different user groups. Or they can follow different crawl logic for collecting their data. But in some circumstances users need to search across multiple profiles.

The new "Metasearch" feature makes it a simple menu-driven procedure for the search administrator. Just designate a master profile that encompasses any two or more sub-profiles. The sub-profiles may even be on separate machines. When the master profile is queried, the search is executed automatically -- and simultaneously -- against each sub profile. The results are integrated into one list, ordered by relevance rank or date. The sub-profiles may be queried separately as desired in other situations.

Access Controls

An enterprise may want to appoint more than one person to administer a Search Appliance. Most commonly, the crawl (indexing) settings for different profiles may need to be set by the individual responsible for each different document collection. Or an organization may want to assign the search settings (user interface) controls separately from the crawl settings.

In one situation we encountered, our customer wanted one user to control which documents to index, but another to control the refresh schedule. In essence, the former is a business user's responsibility, but the latter responsibility belongs to network administration. Both those items appear in the same Walk Settings menu.

The Search Appliance's new Access Control page gives organizations very fine-grained control over which settings are available to which user. There are options to grant or deny permissions to either users or groups; and to grant or deny separately any of approximately 200 different settings and features of the Appliance.

Replication

A new procedure may ease the load on your network resources. Until now, to maintain a redundant Search Appliance, you needed to set the second machine to crawl the same data independently. The replication feature eliminates that step. Instead, the second machine gets a copy of each document or web page from the first machine. Depending on where the second machine is on your network, that has the potential to nearly halve the network resource load of a fully redundant solution.

Index Fields

For every web page or document, the standard search index has traditionally encompassed: title, description, keywords, other meta data, and body text. A new menu under All Walk Settings gives you control over which fields to use, and in which order. You might want to omit document titles if you know they are spurious, for example. Changing the order controls how fields are weighted in evaluating relevancy: the higher in the list, the greater the weight.

Document Level Security.

The Appliance has a group of new features that restrict search results to only the documents a user is authorized to see. Protection methods supported include basic, Windows NTLM, and cookie-based single sign-on.

Selling used books is more complicated than selling new. Consider the inventory: The USA's premier used book dealer, Alibris, Inc., has more than 45 million items for sale in its database!

The centerpiece of Alibris's business is its online inventory database powered by Thunderstone's TEXIS software. The Texis database and search engine have grown along with Alibris, to its role today as USA's biggest bookstore on the web.

There are other aspects besides the sheer number of items, that make used book searching a special challenge in comparison to new. A new book database changes little day to day, as only a few titles are added or subtracted from the master list. Just the item counts need to be updated as copies are sold or shipments received. But in a used bookstore, each physical copy of a book is a separate inventory item. That's because each may be a different edition or in different condition, and must be described separately.

That means a used book inventory is very dynamic. As soon as a customer buys one copy of a book, that listing must be deleted from the database. The search indexes must be constantly updated with additions and deletions. That's an extra burden in comparison to any relatively static text search index. Alibris even has a facility for third-party dealers to add their inventory to the database. The dealers -- thousands of them -- upload their additions or deletions as often as necessary, adding to the dynamism that Texis must accommodate. Alibris gets 3 million update requests a day!

The database's size also makes it critical to offer users a broad range of advanced search options. Alibris's advanced search page has one of the most sophisticated interfaces of all online bookstores, thanks to Texis. Beyond the usual criteria of title, author, publisher, and date, there are many additional choices such as language, binding, and whether it is signed or a first edition.

Text searches also can be limited by numeric ranges -- for example, by setting a lower and upper price, or a beginning and ending publication date. That's a tricky thing to do with other search products, but straightforward for Texis using a compound text and sorted-order index.

The Alibris advanced search page also takes advantage of Texis compound indexes to provide a whole variety of sorting options. Results can be sorted by condition, author, title, publication date, or price -- in up or down order.

"Texis really helps satisfy our most demanding customers, or those doing complicated searches," said Steve Gillan, Chief Financial Officer of Alibris.

Alibris has run its online bookstore on Texis from its earliest days, when it began as a confederation of independent dealers. Today Alibris operates a tremendous warehouse, and fulfills used book orders for Barnes and Noble, Borders, Amazon, and many other distribution partners. Alibris also has become the major used book supplier for libraries around the world.

"Texis has scaled right up with our business growth, both in terms of database size as well as transaction load. And as our business continues to evolve, so do the demands we place on Texis," said Gillan.

"Our retail and wholesale distribution deals have increased the demands on our search engine for specialized searching, as well as for application integration," Gillan added. "To us, a search engine must be more of a solution development platform than a set of indexes. We create new search features on an ongoing basis. With Texis we're quite confident of our ability to handle each new requirement that comes along."

It's finally here! Thunderstone now supports the Macintosh OS X platform for both Texis and Webinator. For your testing, the free Webinator Mac version may be downloaded from our web site.

Notes for Macintosh programmers and system administrators: Only MacOS 10.3 or higher is supported currently. Installation will follow Unix conventions, which may differ from those typical of Mac System 9. In the documentation, where different procedures are noted for Unix vs. Windows environments, follow the Unix version.

The Thunderstone Search Appliance is now available through the Federal Supply Service of the U.S. General Services Administration (GSA). This means government buyers may purchase it "off-the-shelf" at a pre-negotiated price that satisfies competitive bidding requirements. Some state and local governments also have arrangements to purchase through GSA.

The Appliance is listed on the GSA's Information Technology Schedule, also known as Schedule 70. Thunderstone's GSA contract number is GS-35F-0914P. Government buyers should contact Thunderstone's to request our GSA price schedule.

In case you missed it, InfoWorld Magazine Test Center put Thunderstone's Search Appliance though its paces in its Oct. 18 issue... head-to-head against Google's appliance. Both were rated "very good," receiving an identical score of 8.0. The InfoWorld piece comes on the heels of similar test results reported by eWeek in July. See InfoWorld article or eWeek article (links open new browser window).

InfoWorld and eWeek are the two leading publications that operate labs to do hands-on testing of enterprise information technology products. Most other tech industry publications report about products without ever touching them -- relying on interviews and vendor-supplied literature. For this reason, technology executives recognize InfoWorld and eWeek as two of the most authoritative sources for information about enterprise products. Thunderstone is proud to have largely satisfied the expectations of their seasoned reviewers.

Thunderstone will be exhibiting at the FOSE technology exposition, April 5-7 in Washington D.C. This is intended mainly for government technology buyers and contractors, but representatives from private industry are welcomed as well. Thunderstone customers who would like to attend can save $50 using this free pass. Please stop by to say hello at our display, booth 1236!

August is a slow period in many businesses and government agencies. That also may make it a good time to schedule advanced professional education. For that reason we are planning a three-day developer training in Cleveland, Ohio, in August. The planned agenda is:

• Day 1: The "dowalk" crawler and search applications of Webinator and the Search Appliance.
• Day 2: Application development using Texis Web Script (Vortex) and Texis SQL.
• Day 3: Advanced Texis topics such as performance optimization for large-scale applications.

Customers should consider this opportunity to train additional developers or administrators in maintaining their Thunderstone applications. Attendees may sign up for only one or two days as appropriate. Please contact us for further details.