Magalie Roman Salas
Office of the Secretary
Federal Communications Commission
445 12th Street, S.W.
Washington D.S. 20554

Dear Sir --

I am responding to the request for public comments made by the FCC with regard to docket 96-45, a proposal related to the implementation of the "Children's Internet Protection Act (CHIP [sic])," included as part of the Consolidated Appropriations Act, 2001, which, among other things, requires that public libraries and schools which wish to receive discounted "e-rate" Internet access must incorporate so-called "filtering" software in order to restrict the type of content which may be retrieved using library resources.

Below, I will provide my comments on how the use of filtering software itself is highly anti-democratic, unethical, and unconscionable in our supposedly "enlightened" society. However, since the appropriateness of content filtering itself is not at issue here, I will provide my recommendations for the implementation of this misguided policy.

Problems and limitations of content filtering:

• First, while content-filtering itself may be appropriate in a very limited subset of situations, federally mandated content-filtering for Internet access within libraries and schools constitutes an indefensible over-reaching of local authority and local definitions of what may constitute, for example "obscenity," the definition of which varies substantially from one community to another. This attempt of federal legislators to assume control over content selection, which local librarians and school boards have been able to accomplish for years on their own with a high degree of success (as measured by public satisfaction with the content (books, newspapers, etc.) available in public libraries) is highly inappropriate and misguided.

• Second, the law is exceedingly vague. For example, it requires that libraries and schools wishing to obtain discounted Internet access block material which is "harmful to minors," without clearly defining what this means, other than to say, for example, material which "taken as a whole lacks serious literary, artistic, or scientific value to minors" should be blocked. The point which is completely lost here is the significance of context in determining what may or may not be harmful to minors, and this is one of the primary areas in which computer-based content filtering is inadequate, as a piece of computer software cannot adequately distinguish between appropriate and inappropriate contexts.

• Third, filtering software typically works in one of the following three ways, each of which has significant flaws:
  1. the software may block users from accessing any information whose address (URL, domain name, IP address or block, or filename, for example) does not specifically appear on a "white list" of approved material;

  2. the software may allow access to all material whose address (URL, domain name, IP address or block, or filename, for example) is not specifically banned by a "black list"; or, finally,

  3. the software may deny or allow access based upon the presence, number, and/or proximity of certain "key" words to be found by searching the text

  The problems with the first method should be rather self-evident: large quantities of "legitimate" material will become unavailable to library patrons or school students. This may be the result of several reasons: the content may simply not have been indexed by the author of the filtering software and, hence, may never have been provided with the opportunity to receive an "approved" rating and be added to the "white list"; the content at a given URL, IP address or block, or domain name may have changed or been removed since the list was created, thus invalidating any such list; finally, new content may have been added or used to replace existing content, which might change the status of a given URL, domain name, etc. from "approved" to "unapproved" or vice versa. The amount of content on the Internet is constantly increasing and changing at a rapid rate, and this method is clearly unable to cope with this rate of change, even with regular updates.

  The second method also presents similar problems: much so-called (by the software developer) "inappropriate" content may still be available to users simply because it did not exist at the time when the "blocked-material" list was created or because the content at certain locations has changed from "appropriate" to "inappropriate" since the list was created. Similarly, "appropriate" content may be blocked because it was erroneously added to the "blocked-material" list or because it replaced content at a particular location which was previously considered to be "inappropriate" by the software developers.

  The third method is by far the worst filtering method and will result in many "false-positive" and "false-negative" decisions. By filtering on keywords without exercising judgment over context (as only a human librarian or teacher can do reliably), the software is unable to provide anything remotely effective at distinguishing between what should and should not be filtered. For example, software which filters out the string "XXX" would likely prevent legitimate access to material which discusses the sporting event "Super Bowl XXX." Similarly, filters based on the word "breast" would fail to distinguish between material discussing "breast cancer" and "breast pictures".

  All of these methods fail to reliably take into account the context in which the material which may be allowed or denied is sought. Further, the lists used by each method (keywords, allowed-content, and denied-content) are often not made available to the public by the software designers, which provides a significant opportunity for software developers to incorporate a "hidden agenda" of certain political or moral values without public knowledge, and also makes it difficult or impossible for librarians to determine whether a given piece of software is capable of performing the appropriate function, and extending my argument that none of the filtering software packages which is currently available or is likely to become available in the foreseeable future is capable of guaranteeing that all "inappropriate" material will be blocked or that no "appropriate material" will be accidentally blocked, as none of this software can or does take context into account.

• Next, it must be noted that much filtering software which is currently in use can be easily circumvented by users. This is a widely-known fact which the developers of filtering software are unlikely to publicize. Thus, whatever effectiveness such filtering software might have had is significantly diminished if its purpose can be easily defeated.

• Also, as I have mentioned above, it is important to realize that local and not national standards have historically been used to distinguish between, for example, what is "obscen"e and what is not. This applies in many aspects of government; for example, some communities may permit the existence of strip clubs, while others may ban them on the grounds of obscenity. Similarly, some libraries may carry certain magazines or books which others decided would be considered "obscene" by the local community. Since it is unlikely that there would be as many variations on filtering software implementation as there are in community standards for obscenity, there are likely to always be problems with implementing filtering software to match a given community's definition thereof. These problems are made particularly acute because, as noted above, most developers of filtering software do not make public their lists of what their software actually filters and why (which further allows these developers to impose a hidden agenda of what their own definitions of "inappropriate" or "obscene" might be with near-complete impunity as a result of obscurity).

In short, the filtering requirement is inappropriate, as it subverts the local control of libraries and education, oversteps the bounds of federal power, and, most importantly, requires the use of a technology which is not appropriate in all areas and which is known to be seriously flawed.

Further, the "monitoring" requirement for minors' use of the Internet in schools is also a cause for concern. I will not address this in detail here for reasons of brevity, but, suffice to say, that this is rarely necessary in school environments and that it may well end up draining significant technological and personnel resources away from areas which are in much more dire need of same.

Comments on implementation:

Here, I come to the real purpose of this letter. Given that the FCC is charged with the enforcement of this law and that filtering software simply must be implemented in public schools and libraries despite its many flaws, then the implementors (the schools and libraries themselves) should be required to meet the following requirements prior to being granted the reduced-rate access:

• The following information about the filtering software in use should be required to be made part of every school's and library's FCC filings and should also be prominently displayed in the library itself: product name, version number, platform, and, where applicable, the manufacturer's name and contact information, as well as the date and version number of the filtering lists. If multiple versions or pieces of software are used, they must be listed separately.

• The entire contents of the filtering lists themselves and other mechanisms used by the filtering software should be made part of the FCC filing and should be made available to the public as with the above information. Disclosure of this information should be mandatory and users of software for which this information is not available to the public should not be allowed to qualify for the reduced-rate program. Further, the reasons for blocking each item on the blocking list must be made public.

• Filtering lists must be able to be modified by each school or library in order to include or exclude material as deemed to be appropriate by the librarians or teachers or other appropriate authorities. Any modifications to these lists should be made public and part of the FCC filing, as with the above information.

• The users of filtering software must be able to prove that the filtering software and lists will be updated on a regular basis and that any updates and changes will be reported to the FCC and made public locally as well.

• Finally, the libraries and schools must provide assurance that qualified staff members will be available to disable the filtering software for non-minors for the "legitimate research purpose"s for which such disablement is permitted by the provisions of the CHIP law.

Sincerely,

Scott Norwood