Web Matrix: Overview Matrix

Your Internet Service Shopping List

This is the real meat of the Matrix and where it earns its name. For each server described on its own page, the charts below represent a checklist of features and attributes in a unified interface. In addition, various sections of the charts are linked to relevant descriptive or background material located in this document or elsewhere in the collection.

Matrix Keys

In creating these checklists, I wanted to do more than indicate the presence or lack of a feature with a simple check/no-check system. On the other hand, I also wanted to avoid a rating scale from 1-10 or 1-100, simply because I didn't want to judge whether one server's feature was 5% better or worse than another's.

To this end I chose the following 4 element scale, which sorts the evaluations logically (and graphically):

'*' Bullet or Asterisk: An excellent and rather complete implementation of the function. For example, a system with boolean searching that supports complex queries.
'+' Plus Sign: Acceptable functionality, but lacking a robust implementation. Such a system would let you select whether to join multiple keywords with Boolean AND or Boolean OR, but nothing more.
'-' Minus Sign: A poor level of support for the function, typically because it is implemented in terms of other functionality. An example of this would be a system that simply decides whether multiple keywords should be connected by Boolean AND or Boolean OR, regardless of user convenience.
' ' Blank: The function is simply not supported on this server.

Just because one feature is rated with a plus while another is minus, it doesn't represent an absolute judgment of quality; rather, I mean it to indicate a rough comparison of the same feature across servers. If you still feel that a mistake or injustice has been made, please feel free to mail me.

Overview Matrix

Overview Criteria

Catalog or Index Name: The abbreviated name of the Web Index.
Evaluation of the Index: A comprehensive value that reflects the author's evalution of the index for general Internet navigation and locating particular documents or services, on a scale of 1 to 5 Spiders.
Number of Documents: The number of Web, Gopher, FTP, and other documents referenced by this collection. Keep in mind that some subject catalogs keep a short list of hand-picked rich resources, but the effectiveness of a search engine is proportional to the size of its database.
Contains Web and Gopher Links: The index contains resources found on the Web, Gopher, or both.
FTP, UseNet, ListServ, IRC: The index contains pointers to FTP or other well-known types of Internet resources.
Other Databases: The index contains information gathered from sources that are not located on the Internet, such as MedLine, newspaper newswires, or other commercial databases.
Clarity of the Interface: The layout of the search interface and other pages is easy to learn and use. Well-designed services will offer navigation services across the collection, poor services will be disorganized or littered with obscure link icons.
Speed of the Interface: Relative speed of the server's links and download time for the images. Typically reflects how many users connect to the index, the quality of the search software, the speed of the server hardware, and the server's support for load balancing.
Image Download Time: For each service with a logo or imagemap interface, how much it affect the download time? A small logo is acceptable, but large images, numerous icons, and textured backgrounds cost download speed and bandwidth.
Text-only Support: Support for disabling images and text-only browsers means includes functional navigational aids and alternative to the information in images or icons.
HTML Forms Support: Reflects how well the server make efficient use of HTML Forms in the search interface and in feedback links.
Non-Forms Support: Certain browsers do not support HTML Forms, and must rely on simpler search interfaces. If the server supports non-forms searches, how robust and useful is the search engine?
Located in the United States: The Web server or a mirror site is located in the United States (not even Canada), indicating faster and more reliable network access.

Features List

Catalog or Index Name: The abbreviated name of the Web Index.
Subject Index: The information in the index is organized by subject area, typically in a hierarchic tree of information.
Organization: The organization of the links, including number and depth of subjects, as well as cross-references and general ease of use.
Abstracts: Amount of descriptive information that accompanies links.
Searchable Index: The information in the index is stored in a database, which is accessed by entering relevant search criteria, called keywords, and then displayed in a list of links to the desired documents.
Multiple Keywords: Users can expand or restrict databases searches by entering more than one keyword. Additional controls are often necessary for flexible control over the query.
Natural Language: The ability to accept a query in natural language (as if asking a simple question), ignoring common or incidental words (the, is, and) and extracting relevant keywords for the search.
Boolean Searching: For servers that allows Boolean Searching, this field reflects the sophistication of the feature. Many servers automatically join keywords with Boolean AND or Boolean OR, but only Alta Vista and EINet Galaxy supports complex criteria.
Proximity Searching: The ability to select a document based on the proximity of the keywords in the text. As a rule of thumb, this means that documents with incidental keyword matches will be rated lower than others with highly relevant content.
Keyword Phrase Searching: In selecting highly related keywords, it may be desirable to treat them as a single word phrase to encourage the search engine to find them together. For example, using "Bill Gates" as a query will generate better matches than the Boolean "Bill AND Gates". Phrase searching provides much more specific functionality than either Boolean searching or Proximity Searching.
Regular Expressions: A sophisticated method for specifying keyword patterns, using wildcard characters and other matching functions; its generally available on search engines that are based on Perl or grep software.
Substring Searches: This feature represents the ability to enter a complete or partial word and generate matches containing it. Exceptional servers will examine the keywords a user has entered, and identify the appropriate root word to use as a substring search; I commonly refer to this functionality as Root or Suffix Management.
Sorts Search Results: Many search engines will list the result sets in order of calculated relevance, typically listing the best matches toword the top of the results document and degenerating into poorer matches toward the bottom. This feature makes its easy for users to identify and print only the best 10 or 50 matches in a set.
Limits Search Results: Some servers allow the user to specify a maximum number of documents to return, thus providing better response time and a focused result set. Certain servers enforce a maximum number of search results, to lessen the server load or to encourage user subscription.
Richness of Match Descriptions: This value reflects relevant background information for the documents in the result set, such as match quality, file location, file size, file timestamp, or extracted passages. The more information that a service provides, the easier it is to identify useful documents in the match set.
Custom Search Software: The software that performs a search is critical to the speed and functionality of the service. Servers written using Perl, awk, or other simple scripting tools are much slower than custom software solutions written in C or using special database software.
Searches Filenames and URLs: Servers that can search the filenames and locations are useful for locating documents in a particular location or by a particular author. Unfortunately, such searches often interfere with keyword searches because machine or user names may incidentally match search criteria (such as www-genome.wi.mit.edu or forestry.umn.edu).
Searches Summaries and Keywords: This is perhaps the most reliable type of search, because the the search engine examines very specific words or phrases, rather than the incidental text and file locations. Summaries are a good way to assist search software in finding relevant documents, but require administrative work and the establishment of consistent descriptors and vocabulary control standards.
Searches Document Fulltext: The most flexible type of search, this method applies a brute force search the complete content of the documents for possible matches. Although time-consuming and prone to error, fulltext searches can be simplified or focused with tools such as root management or proximity searching.

Matt Slot * fprefect@ambrosiasw.com * 11/22/96